Mikio L. Braun

ML and AI expert

@mikiobraun

@mikiobraun Twitter Memorial

19,827 tweets · 2008–2024 · 1046 threads

2012

Ok, as a researcher I know everything about working under a deadline, but this #fiscalcliff thing is getting pretty ridculous... .
Replying to @cartazio
at least nothing in the traditional press. ;) I remember that I still was in the late 600s in June... .
Replying to @Nico
man, das ist ja mal ein kruder Mix wie ihn Mr. Random nicht besser hätte produzieren können ;)
@muratk3n yes almost like spring. Only that it's probably getting real cold again before March... :( hope you have nice holidays!
Sa: an inch of new snow. Su: rain on frozen ground (a.k.a. slippery deathtrap). Today: 5C and it all has melted. Tomorrow? 20C and sunshine?
Replying to @alexott_en
@Quantisan Pretty exited about this. If you have feature requests for the jblas side, let me know! I'll see what I can do.
RT @alexott_en: just merged pull request from @Quantisan to make Incanter use jblas (by @mikiobraun ) - huge performance boost for matri ...
Replying to @wrede
unless the mayacalypse still strikes. Did anyone put a proper timezone on that event?
@muratk3n I'm glad the move is over... New office is larger, but no carpet and no proper ceiling covers. But gigabit is nice ;)
Replying to @wrede
at least got an email from the billing co. They say I should get my code in the next 48 hours. Well, we'll see.. ;)
We were just about to resubmit a revision to JMLR only to see that the submission page is closed for the holidays. #noooooo #lackofclosure
Ordered my IntelliJ IDEA 12 upgrade in JetBrain's "Doomsday sale". Curious when they'll send me the new license code ;)
So it's -3C again, snow has long melted, but I just heard wind is changing on Christmas and it'll be 5 to 10C! Just as in my childhood!
Somehow trying to finish to paper revisions, end of year accounting, talks, in the midst of our office move preparations.
Replying to @jasonbaldridge
I see. Our new building is already opened. But holidays are drawing near. We'll see ;)
Replying to @jasonbaldridge
so I see you're moving, too ;) In our case, the building is going to be abandoned. We're one of the last to leave. Spooky.
Last night I dreamt I saw one of the first 4K videos on youtube. What a waste of bandwidth, both dreamwise and networkwise!
Ah, also recompiled ATLAS on my Core i7 notebook. You should be seeing 16 GFLOPS single precision @ 2.667 GHz with jblas-1.2.2 now.
Ok, jblas 1.2.2 has been uploaded, closed, and released to central repo. Now waiting for sync to the central central repo being activated ;)
Cleaning up our offices for the big move. Like the German word "Entsorgen" for recycling stuff. Literally "un-worry".
Put a 32 bit Linux on an external harddrive, recompiling ATLAS for next jblas right now. #whatIdoforyouguys
Replying to @tirsen
I made a similar decision to read all the books. That should be enough for a lifetime. ;)
@thinkberg @munterluggauer natürlich, natürlich, wollte deine Leistung nicht schmählern ;)
Replying to @munterluggauer
der @thinkberg baut da Sachen zusammen, die glaubt man nicht, bis man sieht, was man dafür alles tippen musste. ;)
Replying to @karamshuk
I fear the truth is, when it comes to build tools, there is no simple ;)
I won't lie to you, I find Pythons tendency to name packages just like its most important element slightly amusing.
Sifting through my books packing for our office move I come across such hidden gems as "OpenGL Handbook, v1.2". I think they're at v4.3 now.
Replying to @lawrennd
of course. And I also always make sure to use Bayesian terminology when taking a frequentist point of view and vice versa ;)
Replying to @cartazio
@thinkberg yep. A little bit of fortran parsing & stub generating in Ruby and then Java straight to ATLAS.
Reading "Kernel Topic Models" by Hennig et al. confirms my belief that Bayesians model nicely and then loose it all in the approximation.
2 2
Thinking about jblas, the actually amazing thing is that modern processors are good at floating point computation at all.
Wow, on @thinkberg's MacBook Retina, jblas makes roughly 20 GFLOPS of single precision matrix multiplications.
Alright, let's see if I can finally upload jblas to the central maven repo. Build process based on Ruby/C/Java/Ant/Maven. What can go wrong?
RT @alfonsoeromero: This is a must-read for data-scientists: #PLOSCompBio: Ten Simple Rules for the Open Development of Scientific Softw ...
Replying to @xamat
The Steck-Man? Say hello to him from me. We met during his brief stint at Buhmann's lab ;)
@muratk3n absolutely, man. I also found it psychologically advantageous to always have another layer of insulation ready, just in case ;)
Just to keep you up to date on the winter in #Berlin, yesterday we got about 20cm of snow, today it's above 0C just to freeze over tomorrow.
1
Replying to @leonpalafox
Reminds me of the joke: How does a mathematician catch a herd of sheep? He fences himself in and defines it as outside. ;)
RT @peteskomoroch: @gilgul definitely. Now you need to do a network visualization of the viral propagation of your network visualization ...
Replying to @ChrisDiehl
yes, that's right. Or when it gets above 0C for two days, snow starts to melt, and then freezes again... That is the worst case.
Replying to @ChrisDiehl
definitely not California! ;) Not Nebraska either, but when the wind comes from the east, it can get pretty continental.
Replying to @ChrisDiehl
about -5C to -1C, snowed a few inches today. Snow didn't melt, seems it's going to stay around for now.
Replying to @leonpalafox
@muratk3n usually not (at least in Berlin, the west is a different story, though)
Replying to @leonpalafox
@muratk3n and also closer to the continent, so you have hot summers and cold winters especially in the east.
Replying to @leonpalafox
@muratk3n Germany's climate is pretty messed up. In the north you have the sea, in the south the Alps, the east is pretty flat.
@muratk3n @leonpalafox hehe. No, you nerds ;), literally, winter is coming to Berlin. Below 0C for the past few days and now it's snowing.
Replying to @atpassos_ml
yeah, I have Kindle Keyboard. Works like a charm. Paging is a bit slow ;)
Aggressively worded emails to all on the impeding office move and consequences of not taking care of your stuff.
Replying to @leonpalafox
no, got the old Kindle with keyboard, no touch, just buttons. Great stuff, that ;)
So far the best feature of the Kindle: You can read a book at sub-zero temperatures (metric) with your gloves on.
Replying to @peter_c_william
so now I'm writing chunked encoding because python can only receive that, but not send. #batteriesincludedmya$$
Replying to @leonpalafox
I often don't tweet for a few days but still get like 5 new followers, but they seldom stay ;)
And then there was also that paper due tomorrow which needed a complete overhaul. #livinthepostdoclife
Seems I'm instantly loosing those few followers who have added my while I was silent after I started to tweet again... ;)
RT @MMFlint: "Yes, we destroyed the planet. But for a beautiful moment in time, we created a lot of value for shareholders." http://t.co ...
Halfway through major refactoring: "I think this is one of the cases where it would have been nice to have some working unit tests first.";)
Once you start with manifests and asInstanceOf in Scala, you're quickly back in C land. void*, how have I missed you ;)
New business idea: A site where founders can find tech people to help them build their idea. Now if only I had someone to help me with this.
1
Replying to @DRMacIver
If you ever feel lonely, you should try mentioning your follower count or iPad. ;)
Replying to @mdreid
True story: I never drank coffee before I got my first child. Now I'm up to 2-3 cups per day.
1
I really don't get the whole excitment about WhatsApp. There are tons of free XMPP based chat apps out there.
@muratk3n Only in Sachsen... (Buß und Bettag). Seriously, Berlin has the fewest holidays of all German states...
Forgot my laptop at home. Time to grade that thesis that's been sitting on my desk for a month or so...
@muratk3n At least we don't get that 1m of snow over night kind of thing. At least not in the last few winters ;)
People who're going on- and offline on chat all the time just to remind me of their existence.
@muratk3n Didn't lie to you about the winter in Berlin, did I. And this is only the beginning. It won't get much warmer till March... :(
Still in beginning of the McKinsey big data report, but so far it looks much more substanital than that aweful BoozAllen social media study.
2
Replying to @andyrtd
a cab every morning?? #firstworldproblems. Here's a bridge. Now get over it ;)
Replying to @triologmedia
@stackevil Die sind auf der Liste der Firmen, mit denen ich nie wieder was zu tun haben möchte. Nach der Telekom und Gericom ;)
Replying to @sscdotopen
Schön wär's. Zu dem, was Du meinst, komme ich hoffentlich später heute noch... ;)
Gaah, 4kth tweet coming up. Quick, I need to come up with something witty and insightful to celebrate the occasion.
Questioning the usefulness of imposing arbitrary constraints like a specific initial letter for words describing a concept. #bigdata #v
Replying to @helpingtheblind
yes, unsubscribed and switched to RSS as far as I could. Still many server alerts etc.
Replying to @DRMacIver
I've once reviewed my netbook on my blog. Still suprisingly popular. In Turkey.
RT @twimpact: We're currently working on extracting the real-time analysis engine behind our demos into a web service. Stay tuned!
RT @squarecog: OH: "I had to make that chicken look like a duck. It was a thousand lines of hell."
Let's see, how much we're going to see from that 500k+ RTs "four more years" tweet in the subsampled feed... ;)
Nothing like converting between self-written binary formats on a Friday afternoon so that you don't have to run the whole analysis again.
Replying to @superglaze
Not entirely sure... But it seems the audio interface is also compatible with Windows 7.
Replying to @cartazio
yes, I meant it in a hopeful way. And I know it's hard, having been on both sides of the table... ;)
Replying to @cartazio
I was thinking of supervisors who have become managers. They may not be able to follow the details, but they've done stuff.
Replying to @cartazio
yeah, I had to generalize a bit. But still, even a bad supervisor probably has some experience to share.
What is it with Florida anyway that they always seem to have such a hard time counting the votes. #oldworldquestions
Replying to @ogrisel
Man, this whole big data thing is totally getting out of hand, media-wise ;)
Replying to @mpompery
@zeit_geist "smallish hockey stick" I like it. Sounds like the ultimate insult for a Data Scientist. "He was just a smallish hockey stick";)
Replying to @mpompery
@zeit_geist @appcelerator has someone checked the scaling on the axes?
Replying to @zaxtax
I get that ;) But what I mean is that election outcome predictions have always been based on statistics, even before big data.
What's with all that talk about big data and the US election? Statisticians have been doing that kind of analysis for decades.
Ok, now that the election is over I'd like to See a breakdown where those billions of campaign money were spent.
Kinda hard to admit, but Win 8 is the first Windows in a while which has such a unique style I could get that people fall for it.
Replying to @blattnerma
Not sure. I think with modern multicore CPUs and proper threading you should be able minimize lags during normal typing.
Replying to @blattnerma
I gotta admit, I've come to rely on all the niceties of a full blown IDE when coding in Java or Scala ;)
Replying to @kdnuggets
that's why I said "no short-cuts" ;) thank for the link, btw. We've been thinking about putting together a course ourselves...
Replying to @kdnuggets
... need better training (3) eventually people realize that you can't just train people because it's a pretty complex matter...
Replying to @kdnuggets
Not saying you can learn it. It just seems we have the usual cycle: (1) people recognize DS is important... (2) they say we ...
Replying to @josh_wills
Agreed ;) But hadn't Cloudera started to offer Data Scientist training courses?
RT @josh_wills: If only there was a course that taught you how to do it RT @kdnuggets: Best new jobs in America (#3): Data Scientist htt ...
Replying to @josh_wills
@kdnuggets No short-cut to getting the required experience in the field, I fear.
Replying to @NeurIPSConf
@NipsConference yes, I'm sure It's going to be great. Looking forward to attend next year!
Replying to @jasonbaldridge
Even if it eventually does, it'll wake up as soon as you try to leave the room ;)
1
Kabel Deutschland hatte mir gerade geraten, die Bestellung noch mal aufzugeben, weil sie nach 24h noch nicht im System war. #sischer
Seems like I'll be skipping NIPS this year. 18h flight to Reno and 9h jet lag would've turned me into a zombie till mid-January anyway.
Replying to @cartazio
@alung @IgorCarron @trekkinglemon I don't know... We'd have to ask Yann on G+...
Can't still fully grasp how awesome it is to run Windows in a VirtualBox on Linux. I should have done that years ago! #yeahiknow @thinkberg
Cross-continent collaboration is its own form of insomina. I'm going to sleep, but they're already having breakfast.
Within each mature software framework hidden is the original use case when everything was still easy, elegant, and simple.
3 1
Do you know the feeling when you're completing a project and you're all exited and stuff? Well, polishing a paper doesn't do that for me.
RT @margagual: "Much human potential is lost in the race to publish and become group leaders" Postdocs: Life at the Bottleneck http://t. ...
Installing Windows to get around the reboot-into-Windows nuisance. So far, time savings balance is negative...
Newspaper says the 20C drop last week was a very rare event, like once or twice in a lifetime. Well that's a relief. #tweather #berlin
Replying to @izs
@rickasaurus yeah I know, it's not really a security hole, more a nuisance that anyone can trigger that email.
Not really excited about Win 8, but being able to upgrade straight from Win XP is nice. #butwillitwork
Replying to @mpricemitchell
On the desktop web site, click on the wheel on the top right, then "Settings".
Replying to @superglaze
Oh yeah, I don't know what's gotten into them. I also really prefer physical keyboards, but those seem to disappear, too.
Replying to @draglikepull
.@draglikepull I know, it's not really a security hole, but it's still annoying that anyone can trigger that pw request email for you.
Replying to @the_real_jambi
dayum! You know, I grew up in Cologne, a few 100km west means you have 30 mins more sun in the winter. :(
Replying to @smarimc
@edyong209 I know, it's not really a security hole, but it's still annoying that anyone can request the PW reset.
@holadiho Klar, wenn man genug Daten hat, wird alles Big Data. Letztens erklärte man mir, Plagiariusmus sei auch Big Data.
Somehow to me it seems that administration ultimately boils down to generating pieces of text conforming to differing formal constraints.
Replying to @atpassos_ml
I think this is all part of a bigger plot of academia trying to jump on the big data bandwagon.
@muratk3n April if we're unlucky. Naw, just joking, March should be fine. Mid-march, that is ;)
Replying to @roidrage
Not right now, but had a lot of fun a while back with String.substring(). Not exactly a memory leak just unexpected behavior.
Finally reading about tent.io. Trying to figure out whether each user needs his own server or not.
1
Seriously, if I read about the three Vs of Big Data (or 4, 5, 6?) once more, I'll add "vidiocity" to the list.
RT @eminaksehirli: Hey research lovers! Research in Progress has it's own twitter now: @ResearchInP I won't be offended if you unfollow ...
Harddisk saved. Only adding 5GB to the WIndows partition was oddly ineffective, only a few 100MBs free after the resize... . #dontwannaknow
@holadiho angemessener wäre es wohl gewesen zu fragen, ob die Satelliten-Day-Flat gebucht werden soll oder nicht.
Of course, some time around 4am gparted crashes because xfce was a bit too automount-happy. Let's see what's left of the disk...
Can't say I'm surprised that Pheed's Twitter integration has been cut of. Twitter's TOS states clearly that you can't build a 2nd Twitter.
Readying my laptop for the ubuntu 12.10 upgrade. First, use gparted to make some room #ofcourse #whatcouldgowrong
Open source projects who don't use a public code repository or fail to list any of the authors or contributors.
RT @janl: “I feel idempotent today, nothing I do seems to make a difference.” — @thieso2
Replying to @ravi_mohan
I sometimes reply when they try to be funny ("Haaaaaaadoop... God bless you!"), also being funny. They never reply ;)
Replying to @ravi_mohan
don't know how many recruiters still wanted my CV although I told them I wasn't interested. Must be their reward system.
Replying to @ravi_mohan
I know. I wish it were that easy. Obviously there are people who've pitted their whole existence on doing exactly that for money
Replying to @cartazio
yes, it's funny. Sometimes it's so easy to make people happy other times it's impossible. But for them, it's hard to understand.
Replying to @cartazio
I totally agree. If you've seen enough stuff you know there are no simple solutions.
Replying to @cartazio
I have little hope besides people with common sense talking to one another and making an effort to understand each other ;)
Replying to @cartazio
Yes, that's right. But both terms also describe the idea to use automation to make things more efficient.
Replying to @cartazio
BTW, really like how "algorithmic trading" became "programmatic buying". As I said, every community has its own slang.
Seems like weather forecasts were right looks like a warm sunny autumn day. #Berlin #tweather
Media
Replying to @mcinerneyj
I should compile a list of historically fashionable prefixes and suffixes. "Foo 2.0" is definitely on there, too.
As if the mere technical possibility to run queries on heaps of data would tell you how to find something interesting in it. #betterstopnow
Getting a bit tired of Big Anything. Remember the time when everything was Blah 2000? Go Data 2000!
2
RT @IgorBrigadir: @mikiobraun The trucks... I mean, series of tubes carry steam that makes clouds.
So the occasional boot into Win to work on some grant proposal has left me with 34 updates and about 3 reboots so far.
Can't quite believe that the Windows Java installer is still trying to install some Ask/Yahoo/Metasearch search bar on EVERY. SINGLE. UPDATE
Und die Handwerker und Firmen trenden immer noch... . Kann da mal nicht jemand den manual override einlegen? #trendfail
BTW, did they correct the speed of sound to the smaller pressure up there? I don't even know if it's faster or slower. This keeps bugging me
Replying to @mdreid
@SoloGen still, it's between theoretical work and applied work *in science*. True flow into the real-world is hard to measure.
Given the emotional baggage I've accumulated with respect to the piano, it's a wonder I'm still willing to play.
@edoules yep. In particular, using much more complicated methods only result in marginal improvements.
It's not that the simplest methods are always the best on real data. Rather, usefulness and hardness are uncorrelated and you may be lucky.
If you're wondering about the lack of big data related tweets, I'm working on science right now.
Replying to @raistolo
@mlsec ah, there's a train? America, you surprise me over and over again ;)
WTF @kabel_eins, nach zwei neuen Folgen Castle ist erst mal wieder Schluss? #DeutschesFernsehen
1
A friend of my has to transfer from LaGuardia to JFK next week. Any recommendations? He's a bit anxious... ;)
1
@Thomas_Keil @DrKeil Really wonder what those technical reasons might be... Checking who cares enought to do the switch? ;)
@thinkberg Make him stop working on whatever it is right now and give him some data to play with.
Replying to @headius
Still it seems a bit unfair that endorsing you for JRuby should have the same weight as endorsing some fellow researcher for R ;)
LinkedIn asks me whether I want do endorse @headius on JRuby. Endorse? Dude, he wrote the damn thing!
12 2
Replying to @ChrisDiehl
;) at least it feels like I'll reach my personal limit soon. Thank good the deadline is on Monday.
Today more than usual, the art of science seems to consist in turning coffee and processor cycles into densely formatted pdf files.
3 1
New Android G+ App supposedly supports Pages. Bad news: you need to log out and in to switch. Overall pretty underwhelming.
@bastianventhur ;) just observing that matlab-likeness is much less important to me than it used to be a few years ago.
RT @helpingtheblind: Thanks to @karpathy's who2follow@twitter script I found @johnmyleswhite @quantombone @jakehofman @mikiobraun and mo ...
Not saying that I particularly like libs like D3, but it gives you a bit of perspective how things could be done differently than matlab.
Using pylab again after a few years it feels oddly anachronistic that it tries emulates matlab so hard.
Sometimes paper writing is fun, sometimes it's all pain. Seems like this time it's the latter.
Ok, my problem is probably that only two weeks ago, I was on vacation in the Mediterranean sea with 30C and sunshine #sob
Berlin skipping autumn again, going from a comfy 18C to 3C in a matter of days. I miss the mild falls (and winters!) of western Germany :(
Replying to @cartazio
yeah, me to. Given that I've been doing this for 17 years and more now, anything else would've been a disappointment ;)
Every once in a while I hit the "back" button on my Android till I reach the home screen. #unstackin
RT @Conorpope: "Tell me about Twitter, Dad." "Well son, we would piece together the news of the day by reading strangers' decontextified ...
What I long considered one of the most typical uses of TV, taping shows to watch them later, seems to have become deprecated.
Replying to @chl
;) Ich denke, ich werde irgendso eine billig-box mit USB-Anschluss für die Platte nehmen, bei dem das UI nicht total ätzend ist.
Replying to @chl
Deckt sich mit dem was ich über andere Geräte lese... Überkomplexes UI, instabile Software. Show us some love!
The worst thing about programming languages is that eventually you get used to some of the crazier stuff.
Replying to @noelwelsh
yes, that sort of highly state dependent concurrency might be hard to grasp formally.
Replying to @noelwelsh
I wonder is there any theory what "axiomatic set of concurrency primitives" exist and which are equivalent?
Replying to @noelwelsh
yeah, I know. But people often claim to completely reject JVM concurrency ideas where they "just" provide a better abstraction.
Why is it that with any "non-Java concurrency" framework, if you dig deep enough they all rely on Java concurrency features.
David Graeber was right: money might be the best means of exchange, but aren't companies always after genuine relationships with customers?
Looking for a replacement for our DVB-T harddisk receiver. Apparently a dying product category. Only shoddy, half-baked devices available.
Replying to @cartazio
you're welcome, and thanks. Sometimes I think focus is even more important than time or money.
Replying to @cartazio
yes, memory locality is a very real thing, even more so for disk based storage.
Replying to @cartazio
this is something you shouldn't have to be aware of. But I'm glad he did it, so that we don't have to ;)
Replying to @cartazio
I didn't either. I just read that Goto got so fast because he was even considering the TLA buffer. Which is totally crazy because
Replying to @policecar
;) that must be what Twitter's referring to when they proclaim the need for a "consistent user experience."
Still, it's almost impossible to get used to him saying "you, too" and not think "why, I already have that... ."
Whenever he wants something someone else has, my 2.5yo son has started saying "you, too." On some level, this makes perfect sense.
The bus I take to work goes through the most mind boggling 10 minute detour around some construction works. Includes a highway, too.
RT @StephanNoller: RT @jbenno: Recommendation engines are not bad, just badly implemented reads like 'Communism was just badly implement ...
Well, we didn't get any prizes for our pitch, but talked to lots of interesting peoples. It's always nice to actually talk to people.
Browsing around on OpenFlow and Software Defined Networking. @the_real_jambi said this is the new sh#t.
1
Replying to @DataDays
@ProjectAcom In fact, I know at least four data scientists hired in the last few months by Zalando.
How do you call the discussions taking place on Twitter et al during conferences? Second Room? #DataDays12
1
I think it's pretty great, simple to argue about, good convergence properties, generalizations to high dimensional space, etc... ;)
Still pondering about the best way to bore someone to death with math ;-) IDK, construction of the Lebesgue measure?
RT @wallinchristian: Showing ads for washing machines vs LEGO starwars - funky pitch by @twimpact at #datadays12 Real Time Big Data #bol ...
@muratk3n ah, the 80s ;) every now and then I buy one song from my youth for a trip down memory lane..
RT @jbenno: "if you like to be bored to death with math, you should now ask my collegue @mikiobraun " @thinkberg #datadays
Replying to @andyrtd
Alter, ich kann Dir Geschichten erzählen von Kollegen, die Monate auf ihr Geld warten, weil die Formulare immer zurückkommen. ;)
Replying to @andyrtd
ach ich dachte in der freien Wirtschaft gibt man die Tickets bei der Sekretärin ab und gut ist ;)
@moellus nö. Erster Pitch musste ohne geplante Livedemo auskommen. Die Technik. Das übliche.
More #datadays12 than #strataconf on my timeline. I wonder what that says about me, the attendees at those conferences, and/or those confs.
Replying to @markusandrezak
Lots of features, yes, but IMHO they still give quite a different picture than the way humans see things.
IMHO, recommendation is hard because it's hard to make machines understand the qualities of objects, and there's really no way around this.
1
RT @roidrage: Friendly reminder for all you expats living in Berlin and Germany: Wed. 3/10 is a public holiday. Adjust grocery shopping ...
"But isn't that the general state with AI? For years they're saying '5 more years and we'll get there'" Spencer Reiss at #datadays12
1
Katrin Passig asks "what are those people working on recommendation algorithms doing all day?" ;) #datadays12
2 1
I think recommendations outside of your filter bubble work if there still is some partial link which builds a bridge.
Ah well, @bde says algs should have the courage to recommend something new outside the filter bubble. But not how...
RT @mdreid: @mikiobraun Sure… Next you'll be trying to convince me the web isn't made by robot spiders.
OH: "... and they send the data up to the cloud and then down to a data center..." Dude, the data center IS the cloud.
2
RT @hugopintobr: It took me a few lost clients to start to learn & apply the processes to implement this cycle : http://t.co/PoBECtn ...
Replying to @cartazio
what drives me crazy is when you're on a slow mobile connection and each of the redirects takes 5s without giving you anything.
Replying to @the_real_jambi
yes. Matter of fact, we're also one of the Data Pioneers with @twimpact.
Replying to @cartazio
yep. There's also the occasional redirect to bring you to a mobile version of the page. Which is probably also tracked ;)
I wonder what percentage of redirects have the sole purpose of collecting view and click-through statistics.
I think what Microsoft failed to understand is that Klout is not an analytics company, they're a marketing service ;)
1
@muratk3n yeah, he should know. Got VC money from the US, runs his startup from Berlin. Xoogler, too ;)
RT @mariofusco: There is nothing so useless as doing efficiently something that should not be done at all
Replying to @mariofusco
Yeah and nothing as frustrating as being good at something you really don't care about ;)
@muratk3n @taslanous yeah, even for something you really like to do, it'd be usually considered unhealthy to do it 8 hours a day ;)
Replying to @Nico
Ich kenn an der TU Berlin jemanden, der Schreibmaschinen sammelt, soll ich das mal weiterleiten? ;)
RT @taslanous: @mikiobraun the bad thing about ending vacations is that you realize that you will have to wait a bunch of months till yo ...
Replying to @taslanous
don't get me started about the bad things of ending vacations ;) Still sifting through my email inbox...
The nice thing about just returning from your vacations is that you're still aware how odd it is to spend hours sitting at the same desk.
Replying to @mariofusco
I guess the problem is that in Java anything can potentially be a null while in Scala these are at least flagged as Options.
Replying to @mariofusco
Amen, brother ;) I also don't get how people behave as if nulls might just pop up at random.
Replying to @xyzdkfe
@omitevski @leonpalafox You probably should ask before you send drafts, though ;)
Replying to @xyzdkfe
@omitevski @leonpalafox I didn't mean to criticize metaoptimize. But I think students deserve better support during their thesis.
So it seems that you need to have the Amazon App store installed to run apps bought there. Only that's it ~30MB large (why?) Time to root...
Replying to @leonpalafox
poor people, asking (more or less) complete strangers for support in something so important as a thesis...
Replying to @andyrtd
my flight from TXL to Brussels has mostly business suites (and then me, of course ;))
Really on my way to Bristol for the last @PASCALNetwork Steering Committee Meeting. Seems like I'll be flying right into a big rainstorm. :(
Over breakfast Mr. Brown braggs about his brief trip via Brussels to Bristol. "Stop it!" I say briskly, "just bring a brand new umbrella."
@albert_swart I think I only have the "Playground" album. Can you recommend something else?
RT @cartazio: The trap for any small technology biz is leaning on consulting as a revenue center. Locks up manpower and limits reusabili ...
Aaah, you can preorder the Amazon Kindle for Germany. Prices for 7" tablets are seriously going down. #howlongcanIresist
Replying to @jasonbaldridge
Python and Big Data? Seriously? Ah, they also want you to know Java and Hadoop ;)
Replying to @cartazio
@leonpalafox But then, you can't really say the price is ok because the market regulates itself as it's not a normal market.
Replying to @cartazio
@leonpalafox I think that's exactly the point. It's not normal market forces at play, they ride on the academic publishing system.
Replying to @cartazio
@leonpalafox If it weren't required to publish for your scientific career, nobody would use them anymore or pay for their services
@muratk3n Yeah. Saw him live once with the stuff from the Synergy album. That was pretty nice ;)
Replying to @leonpalafox
And Apple is actually doing some creative work. Scientific publishers are just managing people and overseeing typesetting.
Replying to @leonpalafox
I read somewhere that the profit margins in scientific publishing are close to what you could get form drug dealing.
@muratk3n oh yes, especially some of his earlier stuff with Return to Forever and his trio album "Now he sings now he sobs".
It seems I don't have Herbie Hancock's "New Standard" on my phone. What a glaring omission. A wrong to be righted right away.
RT @louisdorard: "Good code is like a good joke, it needs no explanation" says the sticker on @JoliCode 's Benjamin Clay's laptop ;)
If the jury on Apple vs. Samsung really has been misunderstood what they're supposed to do... Priceless... unless you're Samsung ;)
Working in academia acquired skill #11: Turning a dozen or so pages of grant program descriptions into compact mental representations.
Finally met someone who admitted that he never quite understood how to make Hadoop work for real-time.
1
Why is it that every time I have an appointment in the city it gets unbearably hot? Hm. It's probably called summer ;)
@muratk3n if you just need someones email adress to reopen a dormant account, and get access to stored private data, that's bad.
That twittR package is certainly useful - until you run into the request limits, that is.
1
Hm "remind us" might have been a better choice of words. Anyway. Tell me again why I can't re-edit a tweet?
Replying to @alung
@DataSci_Berlin is more like a meetup, with @DataScienceBLN we plan to do data science courses.
Replying to @alung
@DataSci_Berlin @DataScienceBLN Yeah, unrelated. Sorry for the confusion ;)
Replying to @superglaze
That looks nice ;) But then, you boot it up and get Windows. How unsexy is that...
@aaronsw how about a "Tweet" button on TwitterView? How cool would it be if one could retweet a conversation /cc @bradfordcross and @mdreid
1
Replying to @bradfordcross
@mdreid ;) there definitely should be a way to retweet whole conversations at once.
Meh, @masseffect multiplayer disconnects minutes before completing the game three times in a row... .
Replying to @jaykreps
I guess the hardest part is having the courage to say "nobody was good enough today."
RT @munterluggauer: how many comps can answer in earnest: we are taking care of our people, we are treating our customers well? http://t ...
Replying to @tom_wilkie
@zeit_geist There is a proof in the MoMD book on p.127 that you need to store all N bits if you want to count the number of 1's.
Replying to @mpompery
@zeit_geist yes, I like it, too. But I didn't come with it, @tom_wilkie did.
Replying to @tom_wilkie
you're welcome. Also, you're right. But I was exaggerating to get the point across ;)
My Android smartphone is laggier than a ten year old installation of Windows XP right after booting up.
RT @wesmckinn: Deeply saddened by the sudden and tragic loss of a friend and inspirational Python colleague, John Hunter http://t.co/DVA ...
Replying to @jasonbaldridge
I know, right? ;) It's always the same story, people try to quantify something, but then people forget why.
Replying to @andyrtd
ah, you mean that thing where you have a batch processing system and try to do real-time by doing small iterations all the time?
Have heard the argument that optimizing for pageviews or ad clicks doesn't mean you make great websites twice in the last 24 hours.
Replying to @japerk
another case of "it's often the simple stuff which already works good enough"?
@muratk3n sounds like fun ;) it seems hard data on alg trading is extremely hard to come by. Secrecy is probably the biggest asset here.
@muratk3n To tell the truth, wasn't very enlightening. What works is secret, the rest is linear models and a lot of interpretation.
@muratk3n In that case, it was a book on algorithmic trading. After a few recruiters hit me I needed to get a bit of background.
I'm always amazed at the amount of information your brain can take in if you're really interested.
1 1
RT @stoweboyd: What if big data analysis of the Twitter stream is where the biggest value lies? Will they close that down next? http://t ...
Replying to @cartazio
@pablochacin oh yes! And lots of people who depend on the status quo no matter how broken it is, adding to the inertia.
Replying to @cartazio
@pablochacin it's a pretty complex matter and I have no solution either. But saying 'well, that's how it is' ain't no good either.
RT @zeit_geist: the top 4 startup tips: 1.) don't waste time. 2.) know what you are doing. 3.) have enough money. 4.) don't waste time w ...
I wonder how many patents actually make it into products. Otherwise, it's just a lock down on an idea for 20 years.
Somehow, my Gmail is slowly turning into an archive for social network notification messages.
1 1
When exactly have I turned into a repair shop for all the stuff my children destroy (intentionally or unintentionally)?
Replying to @syhw
yeah, but I doubt Apple would've been fine if Samsung had given them proper credit for the inspiration ;)
Replying to @cartazio
easy to defeat, probably, but it'll still be very expensive to do so...
Replying to @cartazio
and people gaming the system to make a lot of money. Another broken system and nobody knows what to do, IMHO ;)
Replying to @cartazio
of course I'm not saying patents are bad per se, but a lot of the current litigation is of little value to the customer
Replying to @cartazio
well I think there are bad kinds of patents. Very broad ones, very obvious ones. And then there patent troll companies, etc.
Replying to @cartazio
I think the problem is that even if you already generalized reasonably, they'll nevertheless try to take it a few steps further.
And if you're lucky and you're patent is accepted as is, the space of "rectangular devices with rounded corners" belongs to you for 20 yrs.
Say you have an idea to do X. Patent lawyers will try to generalize that as much and cover as many variants to do it as possible.
I guess the ruling was right given current law. IMHO the main problem is that you can claim a ridiculously large part of the idea space.
And sometimes, I only remember those auto-completions but not what they really said afterwards. ;)
Sometimes when people pause while telling a story, I'm like "please go on, my brain is producing the weirdest auto-completions."
1
Just saw the first sponsored tweet I found mildly engaging. Perhaps there is hope after all ;)
@muratk3n yep, that's correct. We Germans sometimes do that. Plural for Pizza is Pizzas, for example.
Replying to @cartazio
@atpassos_ml As JFK said: "One person can make a difference and every person should try."
Replying to @cartazio
+1, man, +1. What were they thinking with that new card-based layout on the Android devices... I don't know.
Just found out that thanks to my myopia I'm able to just see individual pixels on an iPhone retina display. #wellthatsgoodtoknow
Replying to @mpompery
@zeit_geist But The Bomb is just a threat. As far as I've understood, they'd like to see algorithmic trading becoming the standard.
Death by logic: if you die after having ignored warning signs because logic told you you're doing the right thing.
Listening to Wayne Krantz's "Long To Be Loose". I find this music oddly addictive. Is also the best showcase for single coil pickups I know.
First time I've heard the argument that there is a push towards algorithmic trading to make markets more stable. What about Knight Capital?
Replying to @atpassos_ml
that and people might have ideas what to change but feel they can't on their own.
Replying to @andyrtd
That's probably how you can make it work for you. But I think their first reaction of WTF? is often correct.
Practically each new Ph.D. student doesn't believe me first when I tell him how core processes like reviewing and public funding work.
1
Why is it that almost everywhere you go, people tell you that the system is broken but that's the way it has to be done?
Forget glass half full / half empty. The question is whether you're a 1 / 0 = ∞ or NaN person.
2
Also, a (non-techie) friend of mine aptly described the Twitter API logo as "a bird in a treadmill".
It's amazing that it's as difficult to explain business value to scientists as it is to describe technical value to business people.
2
RT @josh_wills: @dberkholz @mikiobraun I don't know if academics do quite that much looting and plundering.
Some say academics is like distributed exploration. But how effective is it given little coordination and massive notification delays?
5
Replying to @xyzdkfe
@omitevski @leonpalafox for example by licensing their feeds among them. Have Justin Bieber on your network? Others can get him, too, for $.
Replying to @xyzdkfe
@omitevski @leonpalafox but that's what I'm saying. It wouldn't have to be ad $ if one managed to build an infrastructure of networks.
Replying to @xyzdkfe
@omitevski I've probably been too close to people working in bioinformatics to really believe in that anymore ;)
@radiomaru "we're sex bob-omb and we're here to make you sad and think about death and stuff" #ftw!
Replying to @conikeec
@cloudphysics happy birthday and congrats to your launch! I wish you all the best!
Well, it seems it's already light outside at 5:30. What I would have paid not to know that.
Anyway, don't build just another social network, lay the foundation for an industry. (cc @daltonc )
1
And if you get it right, and people like what you do, you might be able to build a business.
Now if there were an infrastructure in place to hook into and sell your feeds, it would open opportunities to try something new.
Right now there is little innovation in social networks because you can't just test drive your own ideas.
I don't doubt that it's possible technically. But I think we could have a self-sustained ad free industry around social network sites.
Replying to @mpompery
@zeit_geist I don't doubt that. But I think you could have a real industry around that to finally get rid of those ad paid social networks.
So the way to start that would be a protocol plus a reference implementation to get started.
But what I'd really like to see is multiple sites which openly (but not for free) communicate with each other.
RT @mslinn: @omitevski @mikiobraun Fear not! The world will still be here tomorrow, and we'll still be building cool things no-one ever ...
Replying to @sscdotopen
yes, sure. They get a lot of things right, but they too, seem to be trying to be the One Network.
Replying to @xyzdkfe
@omitevski @mslinn Also, just because it's hard to build a successful company, should we just stop? Certainly not.
Replying to @mpompery
@zeit_geist Sometimes I think it's the German entrepreneurial spirit (high quality and sustainable) vs. the American (get rich or die tryin)
@crnixon Good look with your EMACS book! I'm using it since '93 now, and other's should, too ;)
Replying to @mpompery
@zeit_geist But it's of course also insane to plan to become a big company like Google or M$. You need a lot of luck for that to happen.
Replying to @mpompery
@zeit_geist I think it's always hard to build a new company, but that shouldn't keep you from trying ;)
Replying to @debasishg
@thinkberg told me about PSI - Personal Software Inspector which knows about a lot of packages and deals with them in the bg.
RT @boia01: @jamesiry Three UDP packets walk into a bar. Barmaid ask "Now what's going to be your order?"
Replying to @debasishg
absolutely. Microsoft should have built something like an app store years ago.
RT @nik: I come from an alternate reality in which Twitter did not tell everyone these things back at Chirp
Replying to @ikai
I agree. Their stuff looked impressive, but I'm not sure how stable it was.
Replying to @xyzdkfe
@omitevski @leonpalafox I think it's even more severe, because people were trying to build businesses on it, and are then told "sorry".
Replying to @leonpalafox
I think what people are getting angry about is that they used to tell people: "go and build somthing cool", ...
Replying to @leonpalafox
yeah, but Apple didn't started that free. I think they always had a lot of control over everything.
At this point, I'd even be happier if they started charging monthly based on your reach. First 500 followers are free, then log-scale.
One should probably call it the "Twitter model of innovation": start open, acqu-hire companies who build interesting stuff, $$$ exit.
Twitter: Why don't you keep working on the nifty stuff we don't till we can cherry-pick the best ones and lock down on the rest?
1 1
Replying to @sscdotopen
I like it a lot. It's not very advanced but it covers an interesting (& relevant) set of topics.
Let's put it like this, "Mining of Massive Datasets" covers MapReduce in Chapter 2 out of 9.
2
@moellus und da sind ja noch Pläne für'n Startup in der Schublade. Aber erstmal Erfahrung sammeln und Leute kennenlernen und so... .
Replying to @blattnerma
funny thing is, after all this complaining didn't it want to install any updates ;)
@muratk3n Thanks. ;) That my Win partition only has a few 100MB space left doesn't really help either. Oh, did I mention I have to use Word?
It's not that I don't hate Windows, it's just WHY DOES IT HAVE TO UPDATE FOR HOURS EVERY TIME I TURN IT ON.
Booting into Win 7 to work on a grant proposal. I'm sure I'm building up some painful associations here. Question is which one suffers more.
Waiting outside a subway entrance for some downloads to my smartphone to complete. #MobileWeb #2012
I was almost set on the Boss ME-70 guitar effects pedal. Then I saw that @waynekrantz5 demoed the Zoom G5. That changed everything ;)
Writing day today. Project report and updates to a grant proposal due tomorrow. Seems teachers were right: writing skills are essential.
1
Replying to @_onionesque
question is, should ML be talking more about things like concepts, abstraction, perception etc.? ;)
@albert_swart @jasonbaldridge a friend of mine works on automated validation. They do a lot of "model checking" ;)
1
Replying to @jasonbaldridge
ML is often pretty geometric, for example. Separating hyperplanes, convex optimization, etc. Needs a bit of explanation...
Replying to @jasonbaldridge
It's probably not that surprising, but the technical jargon is often considerably different from what a layman would expect.
Replying to @johnmyleswhite
ML gets sometimes described making predictions based on regularity in the data, but the jargon doesn't reflect this.
Replying to @johnmyleswhite
Priors also perform a similar function to regularization in Bayesian inference...
Replying to @johnmyleswhite
what triggered this was that I read about an "Rao-Blackwellized algorithm". I might even get what they mean, but still...
Replying to @muellero
@Avinio thanks, we're currently working on adding more recent data and eventually making it real-time!
Replying to @cartazio
I know people who always sign their reviews on principle. But adding randomness certainly leads to nice theoretical properties ;)
Replying to @cartazio
I'm probably a bit gloomy today, but it's like it's more important to get something published at all than to interact with others.
Replying to @ChrisDiehl
Yeah, it seems everytime you mention Yoda, the bot sends you a quote. ;)
Replying to @superglaze
Exactly. Looking on the screen, peripherial vision should be enough to warn you about lampposts. Works for me ;)
Replying to @superglaze
Yeah, probably. But already in normal operation, walking down the street reading something on the screen should do the trick.
Replying to @superglaze
idk, but don'tsmartphones reduce the field of vision in a similar manner? ;)
Oh, I just Ctrl+C'ed in the terminal running TWIMPACT trends trying to cut-and-paste some text. Sorry 'bout that... .
1
Sometimes I wonder whether all that academic structure like double blind reviews has really helped people talking to one another.
Replying to @johnmyleswhite
also gives a completely new meaning to rebasing and history rewriting ;)
Replying to @johnmyleswhite
Exactly ;) Although most of the work goes into getting the pull request accepted...
Hooked my laptop to the big screen to watch live stream of Japan:USA. Now if only my phone had enough battery left for the 2nd (3rd?) screen
Replying to @johnmyleswhite
actually, there's already a process for that, it only involves a lot of paperwork ;)
OH: "One day you'll be sucked into your monitor if you're so close." - "That would have its advantages."
Replying to @atpassos_ml
Although one could do some pretty sick stuff with implicits. And then there's the problem of fitting it into Scala collections.
Replying to @atpassos_ml
Yeah, I think I'll keep calling it jblas. The Scala code will be more like a thin adapter layer anyway ;)
It seems you can define Scala operators in Java easily. But what should the scala version of jblas be called? sjblas? jblaz? sblas?
Replying to @zoltanvarju
Now that looks interesting! I've been doing some reading into online LDA recently for exactly that.
RT @bradfordcross: Proper comment ranking can turn an uninteresting story into delightful snarktravaganza. http://t.co/Gau2ZfST
Media
People often say that in the industry it's only about money, but I'd rather say it's about people and understanding their needs.
1
RT @isaach: 27 hours 19 minutes 45 seconds later, Hadoop spits out the 12 bytes I was looking for.
RT @ChrisDiehl: +100 RT @gappy3000: If life gives you lemons, make citric acid. Add zinc and copper, and you've got yourself a battery.
@abeatgeek reminds me of when my prof asked me what would happen if we changed X to Y. 'just a sec --- wouldn't work!' ;)
Couldn't one see other people's timeline on Twitter once? Can't find it right now. Curious to see what the world looks to other people.
Hm, maybe saying "to test empirically" is just our way of introducing a bit of redundancy in our language to improve understanding.
1
Replying to @johnmyleswhite
oh yes, I think MLers have a pretty strange notion of what an experiment is ;)
Replying to @_onionesque
Or "let's find a formal proof for that". I know I've used this phrase ;)
Replying to @johnmyleswhite
but aren't most ML experiments numerical simulations of some sort?
@haiqus I get what they mean (and use that phrase myself, I guess). But still, if you think about it ;)
Why do people always talk about "empirically testing something". How could you test something theoretically?
1
I wonder when I'll stop firing up Ruby's irb when I need to calculate something really quick.
@muratk3n I think at least that clickable TrackPad is a typical Apple thing. Don't have a Mac myself... ;)
@muratk3n man, I feel your pain. Is this Acer trying to copy a MacBook Air feature by feature without any sense?
Replying to @_onionesque
Ah, robotics. They can always show nice videos about their work ;)
Replying to @_onionesque
Yes, that's right, irrespective of what the Internet believes, "data science" is hardly as big in mainstream as we'd like. ;)
Replying to @syhw
Have you seen "Margin Call". Junior analyst Seth says he has a "Ph.D. in propulsion from MIT" - "So you're a rocket scientist?" :D
BTW, in German, we don't say "this ain't rocket science". I think the closest expression is "this ain't higher mathematics".
2
Retweet this if you changed to two-step-authentication after reading about how wired's @mat's accounts were hacked. ;)
Ich denke, jetzt ist es Zeit, sich mal wieder über den Regen zu beschweren. #mankannsmiraberauchnichtrechtmachen
If you liked that "Gibbs sampling for the unitiated" paper, make sure to follow @psresnik, one of the authors.
Replying to @noelwelsh
;) I can understand that, too. It's almost like a language barrier between the two subfields. It's a pity...
Replying to @nimalan
I meant for models complex enought such that you need to stick to the more complex methods like variational bayes or Gibbs sampling
Replying to @syhw
Thanks for the link, that looks nice, too. Also, didn't know, videolecture also hosts papers... ;)
Anyone know about a "Complicated Bayesian inference for frequentists" type tutorial? If not, someone should write this. ;)
1
Replying to @leonpalafox
oh, that looks very nice ;) Like the level of detail I was looking for (and seldom found in Bayesian papers...). Thanks!
Replying to @leonpalafox
Yeah, there are a few libraries, but I'd like to at least understand how it's done ;)
Replying to @leonpalafox
yes, but that Griffith/Steyvers paper is pretty inpenetrable unless already know what to do. Do you know a better source?
Reading a few papers to understand how to do inference with LDA. Why are Bayesian methods always so hard to implement??
1
@moellus ich kann bestätigen, dass das keine Idee ist. #wiesoleseichnichtjedenmorgenerstmaltwitter
RT @mdreid: Had fun chatting with @bradfordcross and @aria42. Now even more looking forward to what they have in store for @Prismatic.
That argument by Apple that Samsung replicated the whole user experience is hard to grasp for nerds. Hey, it's iOS vs. Android!
Replying to @ChrisDiehl
I thought "we should write a book called ML baselines" but then realized, these already exist, but not under that name ;)
Replying to @superglaze
if their strength is innovation, it's inevitable that others will follow, but they won't be able to innovate on the same level.
Replying to @superglaze
That Apple stands for innovation and questioning the status quo, and people like Apple because they feel that way, too.
Replying to @superglaze
Have you read "Start with Why"? Somehow, this whole patent stuff doesn't quite align with Apple's supposed WHY.
Replying to @Nico
Bootstrap, ne? Das steht auch schon lange auf meiner Liste für meinen Blog... ;)
@muratk3n one of the problems was that each lang has it's own idea how it manages its objects, so there's a lot of copying going on.
@muratk3n the idea was that you could take matlab objects, plot them using python, do some computation in Java, etc. It was called rhabarber
@muratk3n have I told you I once worked on a new scripting language which would have had superior language interoperability capabilities.
Replying to @superglaze
not bad. Was it really addressed at you or was the mail server a bit too fuzzy with matching real names?
Replying to @nikete
@johnmyleswhite @Thejakeyboy now whoever said you can't have fruitful discussions on Twitter was wrong ;)
Some of the older ML textbooks like Duda & Hart should change their title to "The book of ML baseline method - the stuff that works" ;)
2
Replying to @johnmyleswhite
Recently read it, it's a nice read. IMHO it's one of the better books on the "inspiration" bookshelf.
1
Replying to @johnmyleswhite
Just wondering, any connection between 'why' and that book, 'start with why'?
Replying to @johnmyleswhite
@ChrisDiehl @josephreisinger You mean as part of their research program to understand behavior, or the field itself? ;)
RT @johnmyleswhite: @josephreisinger I completely agree with that. ML is the triumph of asking "small" questions.
RT @ChrisDiehl: +100 RT @josephreisinger: @johnmyleswhite in a sense, ML is an entire field that has given up on 'why' :)
Replying to @ChrisDiehl
@josephreisinger @johnmyleswhite That's probably a bit too harsh, but yes! ;)
Replying to @cartazio
interesting. What exactly are you working on? I like the metric of "performance per math" in a method ;)
Replying to @cartazio
I think mathematicians have an appreciation for elegance and simplicity. But in ML it's often "more math is better".
Replying to @alung
It's a method paper (with application in the "real-world", as we say ;))
RT @syhw: @mikiobraun yup, a classic bias: well explained stuff seems easier and so dumber… :( @leonpalafox
Replying to @leonpalafox
These people also seem to miss the fact that the better a proof is written, the more straightforward it looks.
Replying to @mpompery
@zeit_geist Right ;) I should start writing textbooks on elementary stuff and leave all the interesting stuff as exercises for the reader ;)
1
It seems you cannot publish easy things which work, even if they are new, or just take proven existing methods as building blocks... .
2 3
Replying to @leonpalafox
Yeah, you can see the reviews but not the scores and post a rebuttal if there is some major misunderstanding.
So if you combine existing methods A and B and find a new solution to problem C, that's straightforward and the paper should be rejected?
1 2
2/4 of the elevators in our institute's building seem to be broken, cafeteria shut down. Time to move into the new building.
I know it's hard, but the bulk of the retweet volume on Twitter is about Justin Bieber, a few global events, and retweet games. :(
1
Replying to @sscdotopen
had been thinking about these things for some time now. It was time to write it down ;)
Diese Twitter2Teletext-Bridge ist irgendwie cool. Das Beste aus 30 Jahren Telekommunikation ;) #ard777
Adding open graph metadata to my blog. Perfect job for a Friday afternoon at about 29°C in the office.
Back in office after 4d mini-break during awesome summer weather to look after my kids while the kindergarten is closed. #couldvebeenworse
RT @ntljk: Berlin friends, I do not want to see you on Twitter or the internetz today if you can possibly avoid it. GO OUTSIDE it is gor ...
Replying to @conikeec
yeah, although he's probably a good match to Schwarzenegger's acting skills ;)
Eventually movies will be like classic operas, a handfull of scripts eternally re-imagined, redressed and remade... .
RT @medriscoll: "If you want human real-time interaction, Hadoop MapReduce is not and likely never will be the way to go."- @curtmonash ...
Replying to @whitepapers824
Thanks. Is already fixed in the latest commit. You could try git clone ... and mvn install. But yes, I'll release it soon.
Replying to @superglaze
@thinkberg, who was sitting next to me, just said: "Ah, you don't understand. That's why you're not a real Apple fanboy". ;)
Replying to @superglaze
Especially when you consider how often you will unpack your tablet. IDK, four, three times a day? ;)
As long as Google Analytics won't give you pageviews over the last 24h in a running window, it's hardly truly real-time.
RT @ian_soboroff: Just stop. "Big data" was bad enough. "Data science" bugs me too. But "Data jujitsu"? Some of us are tryna work here.
Replying to @ChrisDiehl
There are things which hurt exploration like people flocking to "hot topics" only, or directions which are considered off limits
Replying to @ChrisDiehl
Good question. I think the level of exploration you get is ok, but it could be guided better.
RT @zeit_geist: @mikiobraun oh yes! apparently marketing killed 'real-time'. The rule of thumb nowadays seems: replace 'cloud', use 'rea ...
Replying to @mpompery
@zeit_geist I've been reading around on real-time business analytics (oh, sorry, "continuous analytics") a bit, and it's even worse there.
Replying to @mpompery
@zeit_geist I had enough after the initial list of 3. There is an insane amount of marketing fluff just like that around right now. :(
Replying to @bradfordcross
@aria42 @jrfinkel @w01fe @keussen Alright, followed all of you. Keep up the great work!
RT @jasonbaldridge: It is hard for those w/out academic research experience to appreciate how tricky evaluation is and how much effort i ...
Replying to @jasonbaldridge
That looks very interesting. And this although all the papers inbetween have without doubt shown how they are better... . :(
RT @jasonbaldridge: @mikiobraun Even w/in academic work, sometimes old methods *beat* the fancier later stuff, e.g. POS induction: http: ...
Replying to @mpompery
@zeit_geist Yeah, I should. Just finishing another post right now... .
I should do a follow-up post: How to trend one million retweets off the Twitter firehose on one laptop.
1
@bradfordcross I'm really impressed with prismatic. So much interesting stuff! And it looks very nice, too!
1
Twitter tip of the day: Don't change your portrait picture, screen name, and name at the same time. People will wonder who the heck you are.
Always nice to see how an I/O intensive process can swap out all your programs on Linux. Forget about CPU load!
Looks like Twitter's finally cleaning up their conversation view in the Web UI. Now you also see multiple replies to a tweet. #yay
@abeatgeek will be cited a lot nevertheless. Not sure whether that's a ";)" or a ":("... ;)
@abeatgeek Yes, I fear it's true. I think they sent out a number of surveys and now they've "storified" the answers. Pretty dull stuff.
Our paper "Quantifying Spatiotemporal Dynamics Of Twitter Replies To News Feeds" has been accepted at the MLSP workshop. CCA FTW!
1 3
That Booz&Co whitepaper looks like a scientific study only the method and data acquisition parts are completely missing.
Replying to @daltonc
yes, already saw that. Also really liked the argument in the pitch video that users will be the customers you want to make happy.
The problem with (online) social networks is that it's very costly to explore that space because you need a few people to test it... .
1
Just had this "I still don't get why Twitter is interesting" talk with a colleague over lunch. Again. #manitslike2012
It's just amazing how much of my work consists in generating paragraphs of text to fill in text-existence-constraints in administration.
RT @JAXenterCOM: The winner is Charlie Nutter (@headius) for his #JRuby work. Congratulations!
@ticking @locks Then you've learned it well ;) But where would you be without the machine model which it has taught you?
RT @UNGlobalPulse: #RealTime = #data available quickly enough to maintain an accurate reflection of reality, such that action may be tak ...
Nice lunch with @zeit_geist discussing academia vs. industry, real-time big data, and how NoSQL was already all there in 1978.
Hm. Infinite redirect loop on Twitter's "connect" page of the web interface. Interesting... .
Replying to @zoltanvarju
Yes, the canvas is a nice and valid way to think about that. IMHO, later chapters (brainstorming, storytelling...) were weaker.
Replying to @zoltanvarju
Yes, mixed feelings for me, too. I've read it quite some time ago, and there was less impact than I had expected.
Replying to @zaxtax
@leonpalafox ah, mutt's still too feature rich. And the configuration has too many options ;)
@leonpalafox ... ond some service where you can store your (encrypted) configs. That would be nice.
Replying to @leonpalafox
Like I said, I think people would be happy with something stripped down, only IMAP, only Unicode, threaded view, etc.
Replying to @mpompery
@zeit_geist I don't think it's dead, there are just some applications where there are better scalability concepts.
Replying to @leonpalafox
I still have a few "real" mail accounts (at the univ. or at our company) Don't want to forward everything to GMail, yet ;)
Replying to @stanfordnlp
yeah. There seem to be some ways to reduce this, but that relates to an independence assumption between parameters, so...
Pulling together a lot of loose ends I've been working on the last ten days. My head's feeling like it's going to explode. #toomuchstate
Replying to @julianeleopold
Ich warte auch noch ab, am Ende stirbt das, bevor es richtig groß wurde, und wo stehe ich dann... ;)
Replying to @noelwelsh
@thinkberg ;) We also had a pretty bad rainstorm yesterday night. But as the sun came out this morning, we were quickly at 28°C.
Replying to @ideanado
No, just making the parser Twitter ready. Stock Lucene doesn't cut it, I fear ;)
Working on our tweet parser I just see the weirdest stuff in people's tweets. Unicode gives people too much freedom to express themselves.
1
@moellus was ist eigentlich mit StudiVZ, der Countdown war doch schon vor Monaten bei null?
No, everything's fine. Wind is coming from the other direction. I wonder why the kids are still asleep... . #famouslastwords
Replying to @dzidorius
I think so. If the counts are accurate, you get quite some exposure. Quite little flow back to my original blog, though.
RT @jaykreps: Trick for productionizing research: read current 3-5 pubs and note the stupid simple thing they all claim to beat, impleme ...
@muratk3n I think it's very cool, but not yet ready for the usual millions of data points AND features set up usually done.
I have to reread about MCMC. Always feels like a bit of magic. How can you sample from the right distribution given so little information?
@muratk3n They propose some ways to decompose that matrix, but that also means introducing an independence assumption between parameters.
Replying to @mdreid
For papers that's fine. For talks some form of embedded viewer would be nice, though. Are there any free ones?
Replying to @mdreid
For me, the biggest surprise was realizing that you can do all that with some sites and JavaScript. No backend programming required!
Replying to @mdreid
Nah, I don't like slideshare neither. Just tried to be a bit objective there ;) Scribd is quite nice so far.
Replying to @leonpalafox
yeah, they discuss some hacks to decompose the matrix which corresponds to independence assumptions between parameters.
After reading the follow-up paper on "Bayesian" stochastic gradient a bit disappointed because you need to invert a #numfeat^2 matrix.
Dear Twitter spammers, I understand you want to get results quickly, but a user with 14 tweets following 1458 people just looks suspicious.
1
@ogrisel @zoltanvarju As I always say, to know how to do that is what distinguishes a data scientist from an academic machine learner.
Replying to @ogrisel
@zoltanvarju I agree. But with the right methods it's at least possible theoretically (and still a lot of work).
Replying to @zoltanvarju
@ogrisel I don't know, I think "big data" is also about methods which can potentially scale to big data. Even if it's python ;)
Replying to @ogrisel
Yeah, you're right, of course. I just find that term "in-memory computations" so ridiculous. ;)
My 5yo daughter said today "People who are dead are lucky." "Why?" "Because every day, someone comes by to decorate your grave."
1
Recruiter mails me about "BPM in Zurich". Ask whether it's a bad sign I think BPM means beats per minute. No reply. They need to lighten up!
Replying to @leonpalafox
They should live-stream it and then directly archive it. What is this? 2012? ;)
Need to learn more about spectral learning. Felix just returned from #ICML2012 and is pretty enthusiastic about that.
Replying to @markusandrezak
At least, if I had a good day, there are some things in the talk where I think "ah, that wasn't too bad". Typically German..
RT @medriscoll: Twitter as an open, publish-subscribe plaform for real-time data feeds? A dream deferred, not lost, imho: http://t.co/LH ...
Ok, wenn jede Auswechselung 5 Minuten dauert => schon mal 15 Minuten von der 2. Halbzeit überstanden... #gerita
Great discussion with @kaggle and @beaucronin on real-world ML. (Thanks to @thinkberg for pasting this together)
Media
1 2
RT @beaucronin: @peteskomoroch @mikiobraun @kaggle Right, so let's give people tools that do everything possible to enable this high-lev ...
RT @peteskomoroch: @mikiobraun @beaucronin @kaggle Data Scientists are a wide spectrum, but the high impact ones write production code & ...
Replying to @beaucronin
@kaggle That's right. But I guess this skill set is what makes data scientists so valuable.
Replying to @kaggle
@beaucronin Yeah, but that's exactly what I'm saying. As a data scientist at a company, you should know how to do that yourself.
RT @kaggle: @beaucronin @mikiobraun I would have thought Kaggle falls more in ML-for-production rather than ML-as-academic. Thoughts?
Replying to @kaggle
@beaucronin So you can stay within your usualy scipy/matlab/R world and don't have to worry about infrastructure.
Replying to @kaggle
@beaucronin I'd say it's for-production because it's solving real problems, an as-academic because you work offline on data files.
IMHO, Google is treating their consumer products like software: refactor ruthlessly. Only problem: the consumer doesn't work at Google.
Replying to @markusandrezak
Couldn't? Don't they decide on the order themselves. Looks like he chickend out in the last second.
Quite a waste of screen space for short text only posts, though. Well, waste of screen estate is at least consistent with the web version ;)
Google completely changed G+ app design, again. Looks a bit like instagram with its square fields per post.
You should be grateful that I'm holding back on retweeting a whole bunch of #bigdata #hype tweets coming in today.
Replying to @leonpalafox
Would also reduce green house emissions a lot because people wouldn't have to fly around the globe all the time ;)
Replying to @leonpalafox
I think this "perpetual conference" is actually possible: Fold the paper, the talk, and the poster discussions into one site.
Now if they would put up the poster/slides as well as the paper through slideshare or scribd, we could have a "perpetual conference" ;)
@msmeissn ah, @thinkberg sagte mir gerade, dass Du mit ihm das telnet-applet geschrieben hast!
Honestly, I'm not yet at the point where I look at d3.js and think "If I started from scratch, I'd end up with something very similar."
1
Now converting a month worth of Twitter analysis snapshot data into the new index format... .
Replying to @alexott_en
most web stores from Lenovo or HP are also so disorganized. I just want to enter some specs and get a list of candidates.
Replying to @fullofscorpionz
@sorenmacbeth I'd rather say that's at most half of it. The real challenge starts where you have to move beyond Hadoop.
RT @marthadear: on the blog: hallo @DB_Bahn, bitte gebt eine echte stellungnahme zu eurer kooperation mit #marschfürdasleben ab! http:// ...
Replying to @sinanaral
No offense, but I probably would if the paper weren't behind a pay-wall.
2
Replying to @superglaze
can't believe I read the whole thing. I probably was waiting for the punch line, but well... ;)
Replying to @leonpalafox
Berlin ;) It's not that bad, just two days of rain after some nice sunny days... .
Replying to @msmeissn
@DB_Bahn Gute Frage... . Aber selbst wenn man vom inhaltlichen abstrahiert fand ich den Ton eher unprofessionell.
Replying to @pablochacin
oh yes. For stuff like C++ template errors that's probably even an improvement. ;)
Nice, @dzone just launched their Big Data / Business Intelligence Zone. And I'm currently the topmost post ;)
RT @danielfinger: Oh, how Iwould love to hear #stevejobs rant about the new Surface "tablets" from #Microsoft!
Replying to @marcmanzano
Yeah, when HP was offloading their tablets for 99$ apiece, people were buying them like crazy ;)
@muratk3n ;) same thing happened to my quora answer on SVMs. Someone came up with a fairy tale about Vapnik which got all the upvotes...
RT @timbenedict: Microsoft had to include a physical keyboard with #surface to give its user access to the CTRL ALT DEL keys.
Replying to @pablochacin
still better than getting stuck with a purely typing issue in Haskell ;)
Replying to @superglaze
Yeah, in particular when it tries to install the Ask/Yahoo/whatever toolbar everytime (at least under Windows)...
Beautiful summer day. Perfect for working on performance improvements for some on-disk indices...
1
Replying to @msmeissn
tehe ;) und ich dachte, die hatten einfach zuviel Unternehmungslust so am späten Sonntag abend...
Was macht man, wenn der ICE schon eine halbe Stunde Verspätung hat? Man schiebt noch einen Halt in Wolfsburg ein... #DieBahn
Just when I thought "who is this? Unfollow!" did I see that it was a sponsored tweet. #verynice
RT @pablochacin: If there were a programmer's hell, it would be written in C++ and hosted by Apache Foundation.
Replying to @cegme
I often have the impression that there isn't a real support team behind Scala, just a bunch of comp lang PhD students ;)
Replying to @cegme
I meant one of the scala people's PhDs, but if this is what it takes to complete your PhD, you have all my support ;)
SAP's Dr. Sikka: "Well you never know, but we have a massive amount of patent coverage on HANA". Now that is nice to know... .
Replying to @cegme
I guess that depends on whether someone's Ph.D. depends on that piece of code working ;)
Replying to @blattnerma
which can be all the world ;) Honestly, I don't care about display dpi, but there are few notebooks which can pack 16GB.
Replying to @weberste
I meant in the sense of "now I want one". I don't care about display dpi too much but I NEED THAT 16GB ;)
What is it with those D3 guys that they always tell me that their plots bobble and wobble? ;)
RT @janl: Dear developers, if you want to see more great speakers at conferences, we need to encourage them to submit to a CfP: http://t ...
Replying to @conikeec
seems to be real, that page you linked to says Bloomberg has also reported it. Still looks a lot like a "me, too" by M$.
Hey @twitter, how about a fast-scroll feature as in the Android contact lists where you can scroll to some hour and date quickly?
Replying to @wrede
I suspect that the lags come from reloading apps between context switches. Although you should have enough RAM to keep it all in mem.
Replying to @JoergM
wow, 1 Tweet, 15k Follower, das nenne ich eine massive Socialmediapräsenz ;)
@wrede Alright, so it's not just me old phone which makes ICS slow... . IMHO, they didn't quite managed to make Android truly snappy.
Replying to @mlsec
have you looked at those products by SAP, Oracle, etc.? Buzzword marketing at its best!
Replying to @ivanku
Yes, that's a nice post. Not that impressive if you compute the tweets per second per core, though ;)
Replying to @mpompery
@zeit_geist Sounds interesting, probably we can chat a bit about our experiences some time.
Replying to @mpompery
@zeit_geist So what's your involvement with real-time big data? You don't work at SAP, do you? ;)
Replying to @mpompery
@zeit_geist Specifically that SAP HANA whitepaper is also written like a patent document. Lack of imagination... ;)
RT @zeit_geist: @mikiobraun yes, in-memory computation is ridiculous (and should die in a fire)
Replying to @mpompery
@zeit_geist So what about compression? You mean that data is compressed in memory?
Replying to @mpompery
@zeit_geist I was reading the SAP HANA whitepaper, which basically looks like an SQL database in memory.
RT @zeit_geist: @mikiobraun it is a buzzword indeed, but actually is not just "store stuff in memory". It's about mechanical sympathy an ...
Replying to @mpompery
@zeit_geist I know that it can get quite complex, like the "best of both worlds". But "in-memory computation" is so generic, I find it funny
Replying to @mpompery
@zeit_geist Ah, hadn't heard the term mechanical sympathy so far. Googling it right now. Looks interesting ;)
Only if you usually think of memory just as disk caches would the concept of "in-memory computations" look exciting and new.
Moments when my smartphone's screen is blank and it's unclear what's taking so long definitely increased after updating to #ICS.
1
#NIPS2012 deadline, then #bbuzz conference, now back to everyday chores & grant proposals. Quite a clash of cultures.
In fact, Sony delivered ICS to their phones as promised. After my experiences with HTC I was a bit sceptical.
Another speaker admitting they were driving blind for years because they didn't have realtime log analysis... ;) @tlossen on Kafka #bbuzz
2 1
Replying to @larsgeorge
very interesting talk. Have you considered stream mining as an initial approximate stage?
"Big data not only about data, but also about scaling analysis at large" Probably I should try to push the term "Big Data Science" more ;)
1
Room suddenly fills up for "from batch to realtime with Hadoop" talk. But... ONE DOES NOT SIMPLY SCALE INTO REALTIME! ;)
1
Replying to @ivanku
yes, totally! Pass on the meme! So many talks here at #bbuz how to scale into real-time with Hadoop...
Alright, posted my slides to all channels. Twitter, tumblr, G+, Facebook, blog... phew... .
RT @YannCluchey: Scalability challenges in "Big Data Science" by @mikiobraun, interesting examples of adapting algorithms to scale or wo ...
RT @Frank_Scholten: "One does not simply scale into real-time" at @mikiobraun's talk #boromir #bbuz #fb http://t.co/UhPF7VbO
Media
Now a short tutorial on the Java Memory Model. Complete with small example shown in eclipse ;) #bbuzz
The @icswm coverage in my timeline is probably a bit over the top, but a fraction of it is what I'd like to see for next #NIPS2012. ;)
Finishing touches on my #bbuzz talk, then off to the Urania. Don't miss my talk at 3:20pm on machine learning and scalability.
5
Talking to one guy in my office in person, chatting with another one, and several others in a group chat. #NIPS2012 deadline.
So here it is, the day of the #NIPS2012 deadline. Time to sing the high song of the deadline, time to make it or break it, etc. etc. ;)
Resisted the urge to bring my guitar to work today so I can work on my fretboard navigation while waiting for experiments to finish.
Replying to @ChrisDiehl
@mdreid oh, yes, you were right about that (at least concerning me ;))
Replying to @ChrisDiehl
@mdreid Twitter induced agreement across three continents, very nice ;) (if I'm not mistaken)
RT @mdreid: @ChrisDiehl @mikiobraun I'd say deadlines are the grandmother of invention since they are the mother of necessity.
Last minute reimplementation of everything for essential speed up. 26 hours till #NIPS2012 deadline...
Replying to @ChrisDiehl
Makes you think what kinds of arguments children have do endure, though. "Do as you're told, or you'll be sick!" ;)
Also finished first draft for the #bbuzz talk on scalability challenges in big data science. 30 Slides might be a bit too much, though.
My daughter sees a bald man and tells me he should be wearing a hat. I say "he probably doesn't like hats." She: "Then have fun being sick!"
Fred Brooks says in "Design of Design" that the waterfall model was only a straw man to argue against. Interesting.
Replying to @volton
It's in Options -> General -> View if I recall correctly. That's in LibreOffice 3.1 and xubuntu with nvidia chipset.
@muratk3n welcome to the north ;) wait till winter, then it's pitch dark at around 4:30pm...
Aha, turning off anti-aliasing in LibreOffice accelerates output by 1500%, renders program suddenly usable. #goodtoknow
1 1
Why is LibreOffice so slow under Linux? It's slower than Google Docs on Firefox 6. It's giving me a gorilla arm by all that virtual weight.
Replying to @beaucronin
oh yes ;) even if they know about math to be able to understand you in princple ;)
Replying to @roidrage
Just joking ;) I still can't get over the fact that "in-memory analytics" has become a hot marketing term used by BI companies.
Replying to @roidrage
It becomes a big data problem if you see the price tag on our in-memory analytics engine ;)
Replying to @Nico
oh ja, bei uns gibt's auch noch den "Bezugserziehertag", da machen die was mit ihrem B.
RT @yarapavan: Ok, Twitter admits mistake of web application architectures and rolls back to a server-side architecture! http://t.co/kCC ...
Time for a math joke: Why are there no uninteresting numbers? If there were, there'd be a smallest one, which would be interesting.
Google's whole social media strategy feels like applying software engineering's refactoring principles to end-user products.
I fear the current summer preview is like the movie trailers: they already showed all the good parts and lied about the ending.
1
Replying to @ChrisDiehl
don't know, they changed it so often. They also have confusingly different levels of rate limiting depending on auth.
Replying to @lawrennd
It was also quite optimistic of you that someone would provide an example on the spot ;)
1
Replying to @lawrennd
Sounds quite exhausting ;) You're always going "man, this is so great" while everyone else is going "nah, what a piece of cr4p..."
1
I'll never get tired of resetting fsc (the fast Scala compile server) between a maven recompile and starting a script. #fixthisalreadypls
Oh great, forgot to reset word histograms after each time bin. Good thing rerun takes less than remaining time till #NIPS2012 deadline.
Replying to @Nico
strahlender Sonnenschein aber frisch... Geht schon, im Schatten könnte es etwas kühl bleiben...
Replying to @headius
Good luck at RedHat! I still remember when you moved to Sun. Keep up the good work!
What's this. The "back" button on my Smartphone is slowly dying. But it's a soft key, how is that even possible.
Turns out one reason a screen session doesn't properly resize its window is that you accidentally started a screen within a screen... .
Replying to @leonpalafox
I see. The media seems to have quite an anti-Facebook bias in Germany then. Which isn't that surprising...
I'm surprised people are shocked that the Facebook stock doesn't perform well. Was there no such thing as the German Telekom IPO in the US?
Always fun to bootstrap your multi-module maven development environment on a new computer. #downloadingtheinternet
Bugger, just when I wanted to fix some bugs for jblas, my @intellijidea OSS license expired with no extension due to lack of updates :(
@muratk3n oh yeah, I'm also relying on some on-disk storage and on-network communication channels ;)
Re-reading the original "Map-Reduce for ML on Multicore" paper by Chu et al. in preparation of my #bbuzz talk.
1 1
RT @josh_wills: Tumblr guy: "Most of our developers are working in PHP every day." Audible gasp of horror from the audience. #hbasecon
Replying to @alung
BTW, I'm also trying to close the gap between practice and theory here. Most people believe you just need some scalable DB, but ...
Replying to @alung
But actually scaling these methods isn't trivial by far. There's a lot of interesting problems to solve... .
Replying to @alung
Yeah, but also in a way which I often find pretty technology averse. For them it's all points in a vector space.
Replying to @alung
And I still have lots of discussions along the lines of "so tell me again why you're on Twitter?" It's 2012, man! ;)
Replying to @alung
Oh, yeah, everyone wants a piece of the cake ;) But IMHO if you're not active in social media, you lack the first hand experience.
Replying to @louisdorard
@regisb Instead, we're losing out to the data mining and computational scientist guys, with a few notable exceptions.
Replying to @louisdorard
@regisb I think for ML it's even more unforgiving because ML is a key technology to analyzing all that data.
Replying to @louisdorard
@regisb Good question. I tend to filter them out after a while ;) (Probably part of the problem), but more web-related ones.
Replying to @louisdorard
@regisb Interesting. There are other conferences during which my timeline is completely swamped with tweets... .
It's awfully quite around the #NIPS2012 hashtag... . Is ML missing the current hype around social media/big data/data science?
1
Replying to @leonpalafox
With Java? All the time. ;) It was just me searching for a bug in maven until I realized it was all my fault ;)
This is where all that "uh Java is so enterprisey"-BS gets you: you always blame the tools first, not yourself #facepalm
Also excellent timing of the nice weather for the last two weeks before the #NIPS2012 deadline.
Seems we're getting another round of summer. After it was barely 10C last week, today it's going to be 28C #berlin #tweather
RT @greenberg: I'm starting a band. I can't really play any instrument or sing, so I'm looking for a few musical co-founders.
Someone on the bus just told me not to post pictures of him on facebook. But I was just browsing. What kind of world have we turned into?
1
RT @benrothke: @klout says @funnyoneliners more influential than @timoreilly. Nonsense! Some Observations on Klout Scores http://t.co/rL ...
Replying to @PFCdgayo
, I'm fully backing you up on this, @AlexanderFurnas and @TheAtlantic did a bad job here!
Replying to @jamesgolick
that's exactly the same page I was also reading right now ;) Interesting!
Replying to @jamesgolick
oh really? That surprises me, too. He probably gave it to Bill Evans during their time together.
Replying to @mariofusco
I think generally people want growth because the population as a whole is increasing. Personally, I'd say stability is fine, too
@moellus Nur damit ich da die zeitliche Abfolge richtig erfasse: die Toiletten waren in einem anderen Stockwerk? ;)
Replying to @jamesgolick
But I guess he often was an important part in applying the final touch, throwing out half the chords, etc.
Replying to @jamesgolick
it's interesting that he actually wrote quite few of the pieces usually associated with him.
Was ist denn da nur bei Hertha-Düsseldorf los??? Zuschauer stürmen 1 Minute vor Ende das Feld...
The BVM #bigdata12 meeting was very interesting. But after standing up at 4:40 this morning, being back in Berlin is also nice ;)
Of course, the whole day I'm sitting inside it's super sunny and now that I have to fly back it's raining in buckets. #aswesayingerman
Replying to @Schoellhammer
der Elefant ist ja auch das Maskottchen von Hadoop... Das Thema zieht sich also durch ;)
Replying to @jbenno
ich dachte es liegt an mir, aber irgendwie ist der Beamer auch unscharf...
Replying to @leonpalafox
@mdreid more like 25 minutes now. Good thing my Ph.D. taught me not to be worried by practical impossibilities...
Alright, we're already running late by 15 minutes and I have the last talk before the next coffee break. After another 4 talks...
Actually wanted to head home early to go to the gym, but this proof is coming together nicely... . Ok, let's do this before the weekend!
Replying to @superglaze
Is anyone keeping count on these things? And is this already Google/Motorola or the original Motorola?
RT @atpassos_ml: @mikiobraun That's what my entire week was about as well, fighting scala versus fighting the whiteboard
@thinkberg It's really strange in particular because the Android app still looks completely different.
Yesterday fighting with Scala and Maven, today proving convergence with bounds'n'stuff. Certianly can't complain about lack of diversity ;)
Upgrading the SD card in my smartphone from 16->32GB. I only copied everything, and now you're telling me everything works? #verysuspicious
Gave @PeerIndex a quick look, but it's the same never-ending stream of "get a higher score by inviting your friends" as everything else...
Replying to @DRMacIver
I'd also subdivide into "unconsciously" and "on purpose". I swear I'd seen a statistics on this question somewhere ;)
1
Replying to @leonpalafox
You mean like in the Matrix? ;) What RSS couldn't do was easy publishing. But you can follow people's stuff, and its open.
The more I think about it, RSS feeds are the mother of social networks. If only Google didn't lobotomize their Reader...
1
In research it's often about beating the competition by a few percent. In practice the simplest solution is usually already good enough.
3 1
Time to set up another module in our @twimpact code base to isolate all that text analysis stuff. #copythepom
1
Made a plan to bring #NIPS2012 deadline, #bbuzz, and work to do on @serienradar in line... . May will be pretty busy, it seems... .
1
Replying to @furukama
@neina_hh @fukami ich würde sagen, ja, weil wir zu blöd sind die komplizierten in 6 Monaten noch zu verstehen ;)
Replying to @Nico
oder Geräte, bei denen es nicht teil des UIs ist, eine glänzende Glasfläche gegen sein Ohr zu drücken.
Replying to @josh_wills
I see ;) Yeah, it's probably a bit more information than you were looking for, then...
Replying to @josh_wills
We get the R stuff straight from CRAN when they are labeled with machine learnig.
@muratk3n What I don't get is that you cannot patent software here, but once you got your patent elsewhere, everything's fine it seems...
Replying to @zoltanvarju
@klabol Yeah, a little too late to prevent the domain name clash, but we did ;)
RT @twimpact: Nun ist serienradar.de endlich online. Trotz Wahlgetwitter ist der Tatort wieder dominierend am Wochenende. Wir... http:// ...
Talk is coming together nicely. Sometimes pieces just fall together and you think: "Oh, this will be nice!" #grateful
1
Replying to @msmeissn
@thinkberg Sony Ericsson: Lange Power-Button drücken, dann gibt's einen Eintrag im Menü. Auf meinem HTC: rooten + screenshot app.
Replying to @msmeissn
@thinkberg Ja, "sudo" ist vielleicht übertrieben. Unter Android allerdings... Da muss man im schlimmsten Fall mit dem debugger ran
I've probably said that before, but the problem with social network sites is that you can't just testdrive an idea.
RT @PFCdgayo: My yesterday's preprint covered by The Physics arXiv Blog by @TechReview: "Twitter Cannot Predict Elections Either" http:/ ...
Motorola should consider doing some good and use their patent portfolio to keep broken and unhealthy products of the markets.
What I don't get is how the German courts are so patent friendly when you cannot patent software in Germany... .
Replying to @PFCdgayo
this looks really, really interesting! Thanks for putting the paper together!
1
Can't quite believe I really used up my monthly 1GB quota of mobile Internet already. 1GB used to be so much...
Looking at Smale & Zhou, 2003, "Estimating the Approximation Error in Learning Theory" and wondering how I did end up there.
I'll post some more information on our #DataSciene seminar soon. Follow @DataScienceBLN for more updates.
Replying to @balajinix
ah I see ;) the talk was videotaped. I'll tweet when the video is online.
Replying to @balajinix
thanks! (This time with the right user mention ;) were you at the talk?
Replying to @leonpalafox
Yes, very nice. Interesting drawing style. Makes me want to visit Japan ;)
One of these days, I gonna buy me a Mac with Omnigraffle and Keynote to escape that latex/beamer/inkscape/libreoffice impress ugliness.
Sprachstandsfestellungsverfahrensdurchführungsverordnung. Yes, German has such long words ;)
2
RT @climagic: LOAD "*",8,1 # This may have been the first command many of you ever typed. RIP Jack Tramiel, who passed yesterday.
I wonder if some of those agent AI people are doing research on simulating online social networks to optimize information flow or something.
Replying to @ryotat
No idea. I think the process is pretty similar, isn't it? Or do ICML reviewers get fewer papers? Then they are double blind.
ICML reviews are of good quality (even when they're not that enthusiastic about our paper). It's not all bad in the academia ;)
Still hate the new Google analytics. The thing I miss most is the overview over my accounts.
Oh great, finally there's an instagram app for android and now facebook buys the whole thing... . #sigh
RT @twimpact: Awesome, our talk at @berlinbuzzwords has been accepted. We will be talking about scalability challenges and big data scie ...
3.4.12: cyclic permutation of 1, 2, 3, 4 day (in Germany). Next: January 4, 2023. #mathjoke
1
Replying to @zoltanvarju
I see. I usually draw the line where people are enrolled in some way at the university.
Replying to @zoltanvarju
definitely. Some people seem to think just because you work at a university, your knowledge can be tapped for free.
Replying to @leonpalafox
exactly ;) Most stories who covered the later story didn't even mention the earlier story (which was pretty big back then).
Replying to @leonpalafox
I have to read the paper to understand the relationship between the studies. Just funny how the media picks up either story.
Twitter may or may not predict movie box office sales. Take your pick. arxiv.org/abs/1003.5699 arxiv.org/abs/1203.4642
Predicting the Future with Social Media
In recent years, social media has become ubiquitous and important for social networking and content sharing. And yet, the content that is generated...
arxiv.org
Why Watching Movie Tweets Won't Tell the Whole Story?
Data from Online Social Networks (OSNs) are providing analysts with an unprecedented access to public opinion on elections, news, movies etc....
arxiv.org
3 3
On my way to @CumberlandLodge for the annual @PASCALNetwork steering committee meeting. Way to early in the morning...
Replying to @DreamFlasher
Yeah, noticed the change of color. Big improvement! I never knew whether I'd already clicked, so I accidentally un+1'd. ;)
I wish they'd designed Android for minimal latency. No number of cores will help against constant reloading of apps.
LibreOffice is certainly compatible with M$ Office in terms of annoyance... . Need to change defaults? Right mouse on object > Edit Styles.
Has G+ received any form of development lately? Still no decent bookmarking, interop with other social networks...
Replying to @mdreid
still remember his talk on the luckiness framework littered with background pics by Britney Spears and Kylie Minogue.
Doing a lot of java.io plumbing to store more than one liblinear-java model in a single file. #ahthefun
Revisiting marge, my little ML Scala framework. Some funny stuff there with the way feature processing can be used on both data or learners.
Skyrim is pretty buggy. This is the second character which ended up with main quests unable to complete. Time go switch to Mass Effect 3 ;)
Just discovered Wayne Krantz. Pretty raw jam session style fusion/jazz/pop crossover in classical guitar trio setup. Very nice.
Replying to @jaykreps
Research is fine but I think they should split the research from the product and only add stuff to the product that is stable.
Replying to @jaykreps
I had the chance to talk to some Scala guys at a conference. It's pretty clear, they're all in there to get Ph.D.s.
Replying to @josh_wills
@thinkberg @codemonkeyism but they probably should stop putting every paper into the next point release.
Replying to @josh_wills
@thinkberg @codemonkeyism Scala's main problem: you don't get Ph.D.s for maintenance and bug fixes.
1
Replying to @ChrisDiehl
Hm. I guess that calls for extensive testing, that is, more math jokes ;)
@thinkberg War für mich auch gar keine Frage. Vor allem war ich es wegen Semesterticket eh immer gewohnt, einfach reinspringen zu können.
Seems like Twitter folk doesn't like math jokes. Lost 10 followers over that pi day joke ;)
Replying to @mariofusco
Arnold Schönhage also used to say: "Never waste a factor of two" ;)
RT @berlinbuzzwords: Thanks for more than 100 submissions for Berlin Buzzwords 2012! #bbuzz - now it's time for our review...
A child's birthday is always also an anniversary of becoming a parent. And of the last time one could get enough sleep ;)
The new iPad might have impressive specs, but I'll better wait to see whether tablets will catch on. Just like this Internet thingy.
2
RT @jasonkincaid: Getting a lot of questions about Tim Cook's pants. They're dark and casual. Think they're denim, reaching out to Apple ...
I wonder how many people are unaware that Google Mail also provides a pretty functional XMPP chat account (with logs, too!)
The sheer number of chat apps (probably just build on top of XMPP) who claim to be "free SMS" by using phone numbers as IDs is staggering.
Replying to @ogrisel
do you know whether pypy does some form of static type inference like V8 to optimize field access?
Clojure in Python... So I know that the JVM is very decent. What about those other JITing VMs? I'm losing track here... .
RT @Nico: Timeline. Timeline. Timeline. Timeline. Timeline. Timeline. Timeline. Timeline. Timeline. Timeline. Timeline. Timeline. Timeli ...
Sitting for about 2 hours on uncomfortable wood chairs waiting for the doctor was probably less helpful for my back than just staying home.
Although you don't need open source for reproducibility (at least not with all the reuse and derived work clauses).
;) RT @thinkberg: confiscating /etc/hosts, sudo rights and android phone from @mikiobraun so he can focus on the ICML paper ...
Usually, the argument is about birds and airplanes and how we've understood aerodynamics and then built something based on those principles.
But inessence, they are saying that you should rather look into the underlying principles instead of trying to rebuild everything in detail.
Hm... It seems that the LeCun/Pereira discussion was only shared to Yann's extended circles. Sorry about that... :(
Apparently, Berlin's mayor Wowereit doesn't have time to meet with Berlin's thriving startup community... . Reality check or ignorance?
Replying to @jonoandre
the big ones (e.g. mongodb) are all fast enough. But the data structure is so different they are not really exchangeable.
Replying to @jonoandre
I think in a framework like Storm your data would be "flowing" through your cluster with little persistence on disk.
Replying to @jonoandre
Oh yes, Hadoop has a completely different set up where you have a bunch of mostly static files on local disks and issue jobs.
Replying to @jonoandre
Apart from that, I think these are very promising ways of scaling such an app.
Replying to @jonoandre
IMHO the main problem with existing stream processing frameworks is that its hard to predict what the max throughput is.
Removing spurious unclosed HTML comment tags from indonesian Wikipedia dump by hand. Ah the joys of the data scientist...
@abeatgeek so I assume that people are looking into that, it's just not that "mainstream" ;) do you have some pointers?
Replying to @ChrisDiehl
... able to generalize? Maybe by inferring word classes or grammatical rules. But English is pretty poor at the latter.
Replying to @ChrisDiehl
I was reading about n-gram smoothing and thinking how you would approach that from an ML point of view. But how would you be ...
For example, n-gram word models won't work well for languages like German with complex verb endings.
1 1
Reading a bit more about NLP I wonder how much the field loses by focusing mostly on English.
1
Tja, nu isser weg. Bei der Gelegenheit ist mir mal aufgefallen, wie groß seine Frau eigentlich ist...
Finally reading about all those n-gram smoothing methods @mlsec had been telling me about. Whoa... .
1
I guess I have to check how many crew members survived the suicide mission in my last playthrough of ME2. I need my whole team! ;)
Sifting through our data to find at least a few of Germans with reasonable retweet volume... .
Crafting a data set to benchmark some tweet language prediction algorithms. CLD is at 8.8% while an aspell dict based predictor makes 5.5%.
Discussing different strategies to reduce locking in our @twimpact backend with @thinkberg. In-memory alone isn't fast enough anymore ;)
1
RT @josh_wills: @peteskomoroch I'm going to release a Hadoop distro that logs so much info you need another Hadoop cluster to store it a ...
Das schlimme ist, dass wir uns so einen Bundespräsidenten ohne größere Konsequenzen für Deutschland leisten können. #amtsbeschädigung
fmt.setTimeZone(TimeZone.getTimeZone("UTC")) is my favorite repetitive use of TimeZone. Can you say TimeZone once more, please ;)
Discussion sessions with Ph.D. students heating up in preparation of the upcoming ICML deadline... .
The kind of career Sebastian Thrun is building is pretty amazing. When I started to study back in 95 he was still at the University of Bonn.
@kred Is there a reason why I have to re-authorize your service with twitter every time I log in?
So I can't update, don't even know whether I really should. And there is no way to get rid of the notification icon either. #sigh
Hm. New update for xperia phones, but only through Windows client. Whose drivers don't install. Also no change logs to be found. #fail
Hm. Has twitter changed the way they encode media in their stream API or is our code just broken?
Replying to @ChrisDiehl
yes indeed. Friday is supposedly the local minimum, getting warmer, er.. , less cold afterwards.
RT @UrbanP1rate: Es werden -20Grad heute und wer heute nacht einen Obdachlosen draussen sieht,der wählt bitte die Nummer vom Kältebus. 0 ...
A markov chain complains "I know how to spell banana, I just don't know when to stop."
13 5
Replying to @DRMacIver
Doesn't bother me. I think all personal blogs should be like this. ;)
RT @PhilosophersSay: Never discourage anyone...who continually makes progress, no matter how slow. - Plato
Replying to @DRMacIver
procrastination where you actually learn somthing potentially useful is simply the best ;)
Replying to @lucaborger
Yes, it look pretty ugly. Publishing a paper in your own journal about papers from your journal based on... you get the idea.
I predict that in the near future, facebook will ask users who don't generate enough page views to be funnier, or else be excluded.
RT @danielerasmus: Absurd argument de jour: Play by our rules or it will be bad for business Facebook's Sandberg Gentry Warns Europe htt ...
Having a hard time wrapping my mind around some collection within collections typing issues in Scala. Damn you, dynamic languages!
RT @business_design: The cover of the Japanese edition of #bmgen is pretty different... wonder how it will be received... http://t.co/6t ...
Replying to @AlexOsterwalder
@business_design yes, for some reason, Japanese like clutter. For example, compare yahoo.jp with yahoo.com.
Seems people are willing to believe whatever fits their preconceptions best. Well, that's what I've been suspecting all along ;)
1 4
The amount of buzz that other twimpact paper gets is pretty astonishing, in particular considering its huge methodoligical flaws.
The built-in dictionaries are a very nice feature of the Kindle. I'm finding myself trying to look up words in normal books as well.
@SenorG do you know about @twimpact? We're doing real-time analysis of Twitter data since 2009 to discover trends.
Thunderbird still sorts emails into the wrong thread occasionally. Time for a "good enough" mail client which is simple and works.
I've been reading several books now on the Kindle. It's very nice, but the layout beyond simple text could often look better.
Replying to @janl
Ansonsten war alles relativ problemlos. Jemand kommt vorbei und baut es ein (kostenlos). Sonst noch kein Kontakt mit dem Service.
Replying to @janl
Im Prinzip schon, aber effektiv sind es dann oft nur so 70-80 MBit/s. Direkt am Modem auch gerne schon mal 94 MBit/s.
Replying to @janl
Hab' Kabel DE 100 seit Oktober. Bislang ohne Probleme. Das Problem ist eher, zuhause dann auch überall Gigabit rumliegen zu haben.
Replying to @Nico
so wie die Jungs, die ihr iPhone an ihren Mac getethert haben - aber nicht wegen Internet.
@holadiho Leute um 9 Uhr einbestellen, auf dem Parkplatz im Auto sitzen, Radio hören und Kaffee trinken bis 9:15 Uhr. #wennichmalchefbin
RT @joestiglitz: Stiglitz: European governments have signed a "mutual suicide pact" by imposing fiscal austerity plans that will collaps ...
Judge attacks Oracle's 'stratospheric' damages claim against Google, postpones trial http://t.co/CKIgrk2K via @engadget
Replying to @matthewhurst
Hm... . That would be third social media analysis platform to go offline after backtype and ellerdale. Any idea why?
When you think about it, Twitter is how a bunch of web designers build a communications company... ;)
@alexismadrigal Just wanted to point you to @twimpact. Founded in 2009, we do real-time analysis of social media streams.
Well, that was an extremely meeting-centric week. No other way to get 2012 started, I guess... . Next week more "real work", hopefully. ;)
Replying to @DRMacIver
Nope, it's just twitter's algorithms for computing those are just very bad.
Replying to @DRMacIver
@thinkberg Yeah, that would be great. I wonder why nobody has thought about this before ;)
Replying to @DRMacIver
Ah well, something else probably also has a better support for conversations ;)
Replying to @DRMacIver
... you get everything twice. I'm also not sure whether that would really be a problem in practice. I haven't tried it...
Replying to @DRMacIver
Ah, I meant for you. So my friend @thinkberg says it's hard to organize people you follow also into private/public and then you..
Replying to @DRMacIver
One of my friends tried 2 accts. The problem is that all your friends would need to do that, too, else you get everything twice.
Replying to @DRMacIver
Hm. Hard to say what exactly is entertaining. Rants about software libraries and your daily commute on the ferries maybe ;)
Replying to @DRMacIver
ah, path.com, like twitter, but it's on smartphones only and private per default. At least that's what I've heard...
Replying to @DRMacIver
That would be a pity. Probably not relevant, but I'm enjoy reading your tweets ;) So, are you going to path as everyone else?
Still getting sucked a lot into the economics literature... . I even forgot how this originally started... .
RT @sinanaral: Reddit will black itself out to protest #SOPA. Their blog says it all: "Stopped they must be; on this all depends." http: ...
RT @twimpact: Not the real TWIMPACT... - You might have stumbled upon the paper “Can Tweets Predict Citations? Metrics of... http://t.co ...
Also, user's who just send the same link via at-replies over and over again should be marked as spammers... @twitter
1
Dear @twitter, how about putting new users in a "noob" state first where at-replies to people who don't follow them back are not propagated?
@holadiho Ich war mal bei einer Theateraufführung von Sartres 'Geschlossene Gesellschaft', wo der ganze Raum (mit Bühne) eine Wartehalle war
Replying to @_onionesque
Don't worry. There are many ways to pressure them into better behavior ;)
Hm. Twitter's asking me a bit too often to re-log in. Wiping session caches frequently, are we?
Browsing through some lectures on distributed optimization. Is it really all network optimization through game theory?
Replying to @wrede
Oha, hab heute noch die Karte formattiert und S2E draufgespielt. Großartig! Danke nochmal für die Ermutigung ;)
Not bad, more than 500MB of memory for apps on an HTC Desire thanks to cyanogenmod and simple2ext. ;)
Every time I have to set up email somewhere I'm amazed how archaic it feels. No shared settings, no conversation view, complex config, etc.
1
Just clicked "mark all as read" on ALL my Google Reader subscriptions. Sent a shiver down my spine. Feels good.
Reorganizing my Google reader subscriptions. Twitter taught me to sift through hundreds of headlines per day. Folders be gone!
Was ist das eigentlich mit diesem "Gesunden neuen Jahr?" Ist das so'ne Berliner Kiste? Da wo ich herkomme, wünscht man sich ein "Frohes"...
@DRMacIver so, just out of curiosity, do you also have an account or delicous, or are you just grabbing the RSS feed?