Mikio L. Braun

ML and AI expert

@mikiobraun

@mikiobraun Twitter Memorial

19,827 tweets · 2008–2024 · 1046 threads

2016

RT @markscott82: 2016 Year in Tech: when the excuse of "we're just a platform" finally fell flat. Worth a read by @cwarzel https://t.co/V7Z…
Replying to @Quesada
I think Trump kinda happened and then it suddenly was just about him anymore.
Here at hackweek, @truemped just lost it because he wrote an http proxy in three lines of clojure ;)
3
Replying to @pjozefak
yes, but as far as I know, they were never involved in day to day operations.
Just realized I'm actually paying more for mobile Internet than landline. And consume more than the cap on my first DSL contract 😅
3
Yay, the amazon music desktop app finally got that rebranding overhaul. #easilypleasedtoday (hi, @IsmailMarmoush)
RT @jllopezpino: @bigdata @mikiobraun Agree! Using microservices & establishing software engineering as a core competency of data scientist…
@hintjens only met you briefly at data2day two years ago. I have been deeply touched by your journey. I wish you safe travels onwards.
RT @hintjens: Tell them I was a writer. A maker of software. A humanist. A father. And many things. But above all, a writer. Thank you. :)
Replying to @heinrichhartman
@andrew_clegg most DS at Zalando also start out with a more research oriented profile but quickly pick up on eng skills.
Replying to @heinrichhartman
yeah. I also think ops is a separate concern. Not all devs are good or willing to do that, too.
So comparing iMessage on iOS 10 vs. Allo one feels like a step ahead and one like a beta technology demo for chatbots. Guess what is what.
1
RT @pickover: Shortest-known paper in a serious math journal. Two sentences. https://t.co/TW2wHn3JPS
Media
Kinda (privately) obsessed with the question whether Swift has potential to become the language of choice for compile-to-binary.
2
Wednesday morning announcement: #iOS10 update went through fine. Thanks for your attention. ;)
3
Replying to @azeem
@caseyjohnston I still remember times where a month wasn't enough time to download a Gig.
Replying to @t_zoeller
thanks. Actually, they wanted to get someone else's statement but he was in a meeting ;)
1
RT @borisontherun: Do you ever wanted 2 have a look behind the scenes of Zalando?Go ahead and enjoy our 1st #AllaboutZ issue!https://t.co/O…
Replying to @agibsonccc
yeah my mother's from Japan. Quite a mind opener to see societies where individualism is not the top priority above all.
Replying to @mpompery
@zeit_geist amen to that. Biggest fallacy around coding imho. "We just need a Wiki/Jira/Mailing List/Slack/Process to fix this" ;)
Replying to @mpompery
@zeit_geist oho. Well as always, the solution to Not Suck This Much is More Communcatoon ;)
Replying to @ctavan
but @DRMacIver is right, too, all this depends highly ob the project and context.
1
Replying to @ctavan
some odd things immediately. Which of course should then be put into more unit or integration tests I guess.
Replying to @ctavan
I think so. Unit tests are fine, but when you say "alright let's try and USE that piece of code" I usually find
Replying to @DRMacIver
at Zalando, the context is usually some microservice within a larger system, so integration tests are rather challenging.
Replying to @DRMacIver
yeah CI should cover a lot of that "check out the branch and run". But then people seem to forget not everything is covered.
@gergrubi which is probably also ok, but it should be clear, but it also needs to be checked IMHO.
@gergrubi turns out, it mostly means "I did a superficial check for my subjective coding style preferences"
@gergrubi that's right. Somehow I was assuming code reviewed means somebody else checked it works.
Replying to @agibsonccc
you could tell yourself there's a super efficient type inference engine in the back ;)
1
Did a little bit of prototyping in Ruby. Still in love with that language 😍 It's odd and quirky but also beautiful. ;)
9
Replying to @iowdah
@DjalelBBZ yes I agree. I meant meta-philosophically I'm enjoying this. Practically it's a pain of course. ;)
1
Replying to @iowdah
@DjalelBBZ I agree. I also really enjoy that there's a thing like meta-overfitting at all.
Replying to @borisontherun
@twitter yup, got them, too. Also, can't block or report them due to "technical difficulties" 🙄
1
Replying to @fs111
well, that's one possibility ;) I don't expect much besides very interesting answers from this question.
Still resisting the urge to make "why did you choose to professionally write software" part of my interview question set.
2
RT @miishke: Machine learning at the peak of hype cycle together with cognitive expert advisers and software-defined anything. https://t.co…
Replying to @miishke
looks like we got a whooole lotta cool stuff coming up. Right after the trough of disillusionment. ;)
RT @DesignSparkRS: If programming languages were exam papers…#FirstSevenLanguages https://t.co/kIdlkTPofm
Media
@MacworldUK hey there. Tried reading an article but I get forwarded to some ad site EVERY FUCKING TIME with no way to go back. 👍👏
Replying to @vambenepe
yes. In terms of public perception it changed things IMHO. But yeah, they probably would have kept on trying.
I sometimes think back to the times when processor cycles were spent on awesome raytracing renderings. And not deeplearning.
2 6
Somehow haven't watched a single olympic contest yet. First I was on vacation and now, well, I guess it's too late anyway...
RT @ZalandoTech: We're hiring PhDs who can design and implement #algorithms to improve our various #machinelearning driven systems: https:/…
Also: happy first Zalando anniversary to meee! #Zanniversery Time sure went by quickly!
2 9
The two weeks I was on vacation and stuck mostly to Twitter. It's alive! Also consumed around 1.5GB of mobile data 🙄
@aplrst yeah, to be honest, already with dato it took a while for me to make the connection.
Replying to @ChengSoonOng
you need the data? Integration is too hard to do generically? Full alignment productwise?
Replying to @ChengSoonOng
I wonder what the lesson is that apparently ML heavy teams can only add value in the context of a large company.
Replying to @ChengSoonOng
that and it shows how inherently hard it is to do business with generic ML solutions.
2 2
Replying to @ChengSoonOng
then again, I never considered buying their services, so I guess I'm partly to blame, too ;)
Replying to @ChengSoonOng
I've seen them at conferences again and again. They always had nice demos. ;)
Finally realized that Turi was Dato!! Met the guys at StrataConf in May. Looks like a somewhat spontaneous acquisition ;)
1 7
@holadiho also ich follow dich seit dem Tweet von Dir wo du dich von nem Kindergeburtstag auf ne Bank zurückgezogen hattest. #instafollow
1
RT @NandoDF: Neuralise (verb,#neuralize): to implement a known thing with deep nets. Usage: Let's neuralize warping, neuralize this! And tr…
Replying to @beaucronin
it's funny because devs always propose technological solutions to communication problems. ;)
Replying to @torbenbrodt
dont get me started about @klout. They were only good at onboarding companies to pay for whatever it is they compute.
1
Idea: train a model to predict what people in your timeline retweet, apply that to the firehose => more content than you can possibly read.
1 2
Replying to @gear_homie
to Googlers, hardware is always the weakest link, getting stronger with the next server upgrade.
Replying to @gruebe
I know the original book. I liked the approach but have yet to see a practical proof of concept.
Replying to @Mandar_2812
I was also thinking about how programming has become all about plugging one framework on top of another.
Hey just realized you could make a faux chinese writing systems with emoji: 💾🐙 🏃🏻 🌍⚡️(Google in its way to world domination).
RT @ClassicFM: Musicians have been sending us their interpretations of 'Cameron's Lament' and they're frankly amazing: https://t.co/KZAOUSv…
Replying to @jasonbaldridge
@haldaume3 neither did I, but tbh, I'm not sure whether I would take the time to learn emacs nowadays...
2
TBH somehow I missed the whole Britain's got a new prime minister thing. I read about Cameron's 🎼 do doo doo, though. ;)
@gappy3000 same here! I started thinking of my inbox as a Twitter timeline recently. Just fav the good stuff for later!! ;)
1
Replying to @PhilDarnowsky
also "I don't want to see this ad anymore BECAUSE I HAVE SEEN IT EVERY FRIGGIN DAY THE PAST WEEK"
1
Replying to @gear_homie
that's like instead of figuring out how to tune something, you make it configurable #pcgaming
2
Replying to @gear_homie
hurhur. Thankfully Android has endless tools for that. #moretoolsisthesolution
Back in the days, terminal really meant terminal and we didn't even have a mouse 👴🏻
2
RT @ThePracticalDev: Can you even call yourself a software developer anymore? https://t.co/4jjdoe5VdT
Media
Replying to @DmitryKan
yes, more like the latter ;) Usually takes some experience to look beyond fancy technology.
Replying to @DmitryKan
Maybe a misunderstanding, I saw it as outgrowing technicalities and focussing on impact. What did you think it was about? ;)
RT @soumithchintala: this year's ICML was buzzing with grad students planning startups as a ticket to quick money rather than to solve some…
Replying to @PhilDarnowsky
I also heard people say that a lot of the libraries is more or less a Java port
Replying to @PhilDarnowsky
hey I'm still looking for a nice compiled to bare metal lang and I thought about trying Erlang #speakingofabsurdlangchoices
1
Hey I'm not a front-end guy but sometimes JavaScript seems to require to make up for its lack of rigidity with some serious dogmatism.
3
Back in Berlin with a (now truly) summerly 22C in the morning, forecast is 28C for today, even 32C for tomorrow. 😎☀️
Found a bar in Helsinki, bunch of Americans at the next table, and for the love of God I have no idea what their background is.
1
Replying to @alung
I only saw Princess Mononoke but would well say it was trippy enough. ;)
1
RT @AnnaPawlicka: First ad on Twitter that I don't mind at all 😃 https://t.co/kkpn0hblC2
Media
RT @jonmsutton: Sometimes I regret teaching my children an evidence-based approach to life #FathersDay https://t.co/adV9FFewIA
Media
RT @gear_homie: Trying out the Yahoo #Aviate Launcher on my #chinachallenge #android in memoriam Yahoo.
Honestly, I just realized that I'm reading my email as if it was a Twitter stream. #3853unreademails
3
Replying to @Springcoil
totally. Would have to be a mixed hardware/software startup, though. Existing technology is just the WORST ;)
RT @gear_homie: Living with UX failures - Automating yourself - Day 6 - China Challenge - I have now been using my 100$... https://t.co/qUz…
My former professor kindly agreed to me to publishing my lecture slides on scalable ML. Now I only have to find an agreement with myself ;)
9
RT @mcs1408: Had an amazing time with the @ZalandoTech crew at #bbuzz . See you next time! https://t.co/fibYlR3ZZI
Media
Replying to @helicalinsight
Um, how about no? Sorry guys, this is not how you do marketing on social media. I'll have to block you.
Replying to @karpathy
I guess you just gotta get that name on that deep learning paper nowadays? ;)
2
Replying to @beaucronin
so either you overestimate my level... or that's just a looot of money ;)
Replying to @agibsonccc
oh yes. TensorFlow supports LSTMs AFAIK but yes, it would have to be a CHAT BOT FRAMEWORK.
2 3
Replying to @agibsonccc
also, I would guess anything related to LSTMs would be muuch more hip right now.
Replying to @Tagesspiegel
hey, was habt ihr denn für superaggro werbung geschaltet auf mobile? Gameabo, 4.99€ für sieben Tage??? #90errufenan
Replying to @alexott_en
yes. Let us try to find a spot. I'm already pretty stocked with meetings, but we'll figure out something.
RT @hugo_larochelle: TFW you bid on NIPS papers and realize someone submitted one on an idea you wanted to try (courtesy of @awiltsch) http…
RT @BigDataGal: Data! Data! Data! he cried impatiently. "I can't make bricks without clay!" - Sherlock Holmes
So this week has been TXL ✈️ FRA 🚅 Heidelberg for #data2day PC meeting and back, then TXL ✈️ FRA ✈️ LCY to #StrataHadoop and back. And now 😴
1
Great meeting @bigdata, @ogrisel, @Springcoil, @martingoodson, @twiecki, @StephanEwen, @jbenno and many others in person at #StrataHadoop!
2 6
Aaalmost done with my blog port to bootstrap. Now if I can only merge those three different footer snippets...
Media
Demo of Spark Notebool by @noootsab. You shouldn't be afraid of the JavaScript console ;) #StrataHadoop
Media
2
RT @strataconf: "We're all establishing standards & protocols for how the world thinks about & uses data." @triciawang #StrataHadoop https:…
Nice talk by @twiecki on in-sample vs out-of-sample performance of trading strategies. #StrataHadoop
Media
1 4
Although strictly speaking shouldn't the metaphor somehow involve elephants and whales? ;) #StrataHadoop
Media
1
"This is the slide you might want to take a photo of" @gwenshap on reliability guarantees with Kafka #StrataHadoop
Media
25 41
Alright, London. Yesterday's Hardcore Data Science track was very interesting, let's see what the state of Big Data is... #StrataHadoop
1 2
RT @dataJujitsu: .@mikiobraun nails the difference between developers and data scientists #StrataHadoop https://t.co/vwHvOZnYLM
Media
RT @madsh: Great talk on the work of putting a single formula into production. @strataconf @mikiobraun https://t.co/XJlEXprkMO
Media
Piotr from DeepMind about deep learning for text. This promises to be interesting. #StrataHadoop
Media
8
Very nice talk by Danny Bickson giving an overview of recommender systems. #StrataHadoop #HardcoreDataScience
Media
2
On my way to tomorrow's hardcore data science track at #StrataHadoop in London... being delayed due to bad weather around London.
1
@SpectralFilter yeah. I sometimes go like "there's no way there's honestly that amount of information in there. Ever"
Sometimes, entering a deutsche bahn train on a Sunday night is like stepping into a bar fight. You are just like what the eff happened?
1
Maybe I should try to learn a bit of actual rocket science and then look for opportunities to apply it in my work.
2
Replying to @pavlobaron
it could've totally become something like an alternative high five!!
52 minutes in, and we've just set up the repo ;'D @oldJavaGuy @borovikov @nanwang_t @IsmailMarmoush #inhousehackathon
1
Replying to @philgyford
and I was wondering all the time what those goggles in the profile pic were for 😜
Replying to @PhilDarnowsky
although its not strictly the 70s that are at fault here eventually... .
1
Replying to @xyzdkfe
@omitevski totally. And I think they've made the math based on promises they got. But if it doesn't pay off, the money will be gone, too.
Replying to @xyzdkfe
@omitevski but collectively behaving as if AI is solved is dangerous IMHO.
Replying to @xyzdkfe
@omitevski more conscious and honest about what we promise. As long as you stick to clear areas where ML works I don't see an issue.
Replying to @xyzdkfe
@omitevski no we certainly aren't. But they are doing an investment on stuff we have promised, and all I am saying is that we should be...
Replying to @xyzdkfe
@omitevski that for each and every company which hired data scientists, it is clear beyond doubt that they added massive value.
Replying to @xyzdkfe
@omitevski I know it's very hard to have such a discussion on Twitter. But while there has been breakthrough in some areas, I doubt...
Replying to @xyzdkfe
@omitevski yes, but at some point people will ask whether it really has been worth it, did data scientists really create more value?
2
Replying to @xyzdkfe
@omitevski and don't get me wrong, I'm totally benefiting from this, too. That's why I'm also concerned not to "overheat" this much.
Replying to @xyzdkfe
@omitevski so now it's either you're with us or against us? ;) Also, every bubble bursts, but there's still progress.
Replying to @xyzdkfe
@omitevski do you have a link? And what is the connection to anti-intellectualists?
Replying to @xyzdkfe
@omitevski if by positive over all you mean strongly suggesting to everyone you need to switch to big data technology I agree ;)
1
Replying to @xyzdkfe
@omitevski but yeah. I'm the last to stop this, obviously. But I also don't share the unfettered enthusiasm of som.
Replying to @xyzdkfe
@omitevski yeah but based on hugely inflated assumptions. I mean all this talk about "AI" with no discussion of what AI is or isn't.
Replying to @xyzdkfe
@omitevski let's ride this wave and see whether we can secure a spot in the sunlight.
Replying to @xyzdkfe
@omitevski so what I think is happening is that suddenly there is all this attention around deep learning and people are like...
Replying to @xyzdkfe
@omitevski the great achievements we obviously have had, instead of all this "AI is nigh" talk.
Replying to @xyzdkfe
@omitevski it is more like "OK, this is it??" and then it might be over real quicl. What I'm missing is a more honest communication about..
1
Replying to @xyzdkfe
@omitevski I think the thing that bugs me is that mainstream media is now like "whoa AI is here" but once that stuff gets launched...
Replying to @xyzdkfe
@omitevski I also wouldn't go as far as calling it a conspiracy either ;)
RT @robertoglezcano: What developers are really saying by @ValaAfshar #success #optimism #growthmindset #designthinking https://t.co/U7jdL2…
Replying to @xyzdkfe
@omitevski ever since that "what do neural networks dream" posts there has been a strong push to tell people that "AI is here" by Google.
1 1
Replying to @xyzdkfe
@omitevski I'm happy they finally get the spotlight but I doubt that it is just about this.
Replying to @PaulTurnerWG
@skyviewPaul @holgermu that'll be post-nuclear winter, right? I mean after humanity has lost to SkyNet.
1 1
Replying to @hpgrahsl
to me it seems a lot as if it's the latest handle to brand yourself as innovative. And I (still) care too much about AI it seems.
1
Replying to @hpgrahsl
and I think fundamental problems are yet unsolved. It's way too early to declare victory.
Replying to @hpgrahsl
I was mostly referring to Google and friends rebranding as "AI first" and similiar PR stunts.
1
Alright, think about this, would you actually send back a fitting but AI-suggested canned reply to a friend?
Friend was superpsyched he could buy a decent gfx card on business expenses for gpu computing. Then he learned they don't come with vid out.
4 5
Replying to @vielmettialt
@vielmetti well nowadays you'd need to convert that awk code into a deep network, of course #AIfirst
1
Replying to @vielmettialt
@vielmetti I also meant almighty in the sense of all the systems it can control. Of course it will fail.
Replying to @vielmettialt
@vielmetti but then again if all the devices team up on you, things can also get messy ;)
Replying to @vielmettialt
@vielmetti it's probably slightly safer when you don't have to fight an almighty house AI.
RT @geoffreyamoore: Could we put a moratorium on all discussion of self-driving cars until someone demonstrates a self-starting tele-meetin…
Replying to @miishke
the other day I saw a sign on some electrical installation box saying "whoever changes something here knows what he is doing" ;)
1
Now wait a minute... When and why have cogwheels become the universal icon for settings and configuration?? ⚙🤔 #itsquiteastretchright
2
Replying to @DRMacIver
yeah. Weird how the main OS has just become the host to run whatever (usually Linux based) container and the GUI.
RT @fogus: Nesting syntax ranked by aerodynamic potential: < > ( ) { } [ ] | | begin end
Software frameworks so complex you need to write a book to explain how to use them. #2016
1 2
Hey ho, my prof got features together with @smolix and B. Schölkopf as "digital pioneers" in a Berlin newspaper.
Media
3 5
RT @ZalandoTech: Your team has the right to take reasonable risks to achieve greatness #TechConstitution #ZalandoTech
Replying to @PhilDarnowsky
yeah, I still like it even after all those years. In spite of everyone complaining it is dead.
1
Replying to @breuderink
my thesis was all about looking at scalar products between eigenvectors and the vector of all Ys. There's also a JMLR paper
Replying to @5olarpunk
@wirereporter oh my. I hope they dont have another product which is called BSTRD. ;)
I have the strong suspicion that the click from a mac's touchpad is actually not triggered mechanically but by software.
1 2
Replying to @mscdexdotexe
oh, another build tool I haven't heard of yet ;) (and admittedly I have no knowledge of any depth in the matter)
Replying to @eigentourist
while we're at it, can someone PLEASE unix-ify the JVM build toolchain, too??
iMessage couldn't deliver a message. So it sent it as MMS instead. Which got delivered by email. #dontletdevsfixit
1
RT @ChengSoonOng: Let's quickly write some scripts to do this analysis. No need for version control. #famouslastsciencewords
@vosmann_ go to @ThePracticalDev's profile and check out the other covers, they're all great!
Replying to @nanwang_t
hehe. I meant memory to install space. So many underpowered devices that have too little mem to hold even Google apps updates.
Judging from friends and family, Android's number one headache is insufficient app space.
1 3
Cited Star Wars in a team meeting to have them think about what they really want: ✔️ #searchyourfeelings
4
RT @ThePracticalDev: Does your code depend on fewer than 72 NPM packages? Your code is lonely and you need to read this book. https://t.co/…
RT @AcademicsSay: I don't procrastinate. I proactively accumulate guilt-contingent motivation.
Replying to @ANotowska
not even that but they also expect you to review loads of papers for free. And do editorial work.
RT @ThePracticalDev: The only correct way to write JavaScript is whatever you were not doing last week. https://t.co/p48RgucO3x
Media
Tried to explain to my son that sending videos consumes a lot of data. He: "What is data?" Me: "That is a very good question" 😊
7 19
RT @ZalandoTech: As it's #WorldPoetryDay we wrote a haiku. Zalando Tech rocks Radical Agility The best team ever.
Replying to @beaucronin
have you also been reading these articles about manging millenials? ;)
Replying to @fhueske
well at least they did an update. I heard w10 is finally coming to older devices. Let's wait and see..
Replying to @fhueske
oh there was an update? I didn't notice any change... Most of the newer twitter features are still missing...
Man, I don't get why the Windows Phone Twitter app has to be this bad. Instagram, Vine and Whatsapp show it can be done. #Imtheonepercent
5
RT @robertoglezcano: The evolution of intellectual freedom by @PHDcomics #science #research #researchers #researchpaper #edchat #phdchat h…
RT @tomhenrich: “This election is gonna get pretty interesting.” ‘Define “interesting.”’ “Oh god, oh god, we’re all gonna die?”
Replying to @xamat
@PopSci @taboola actually I wouldn't mind that kind of (hopefully data driven) hyperlinking if the content were legit.
1
Replying to @aCraigPfeifer
Yesyes. There are ways to approximate offline but I fear that's confidential.
1
Replying to @aCraigPfeifer
that I'm actually working on recommendations for retargeting now doesn't make it any better at all ;)
1 2
I already lost to computer programs in the 90s, when they were little more than branch & bound algorithms written in a thousand lines of C.
4
Replying to @benhamner
a friend of mine once said it's not even that but more like a sequence of binary patches.
RT @ZalandoTech: Tomorrow: #AMA with SVP of #Tech Philipp Erler & VP of #Engineering @ebowman about #radicalagility. Send us your questions…
RT @randal_olson: The last #programming book you'll ever need. h/t @ThePracticalDev https://t.co/aQh95nr2bi
Media
Replying to @StephanNoller
@function hey ho, @derMicha und @thinkberg ist auch dabei! Glückwunsch und viel Erfolg!!
2
RT @thecoderlife: When you are a good back end #developer.... #coderlife https://t.co/hvM0Dy9rg8
Media
Replying to @pavlobaron
hehe. Well better to spend a few litres on a case than burning a few tons in the atmosphere I guess.
Replying to @pavlobaron
I think once we run out of oil business travels will become much more infrequent.
Sometimes I think a dev's main skill is to decide after 2h of Internet research which 2m old open source project to you use for the next 2y.
9 17
RT @beaucronin: @mikiobraun another way to look at it: it is no longer possible to have another old-style AI winter. Hype can deflate, but …
Replying to @beaucronin
yeah but I think you are right. Money means there's suddenly a demand for such kind of expertise.
1
Replying to @debasishg
(a) and (b) are definitely true. (c) has been around for some time. I think availability of software is also a key here.
Concerning deep learning, I keep asking myself what has changed so that it became this popular.
1 2
RT @darleysam: It is the straight-faced determination with which a computer tries to struggle through these problems that amuses me https:/…
RT @JonathanAFrye: "Doesn't Work in Science vs Works in Science" - by @twisteddoodles ht @PoliteHunk & @realscientists https://t.co/9TrwZn…
RT @random_walker: 2 decades after DoubleClick founded, % of online ad revenue from behavioral is still *single digits*. Huge failure of mo…
Replying to @fs111
there once was a recruiter who sent me the exact same mail every few weeks. Turns out he was human, just very unorganized.
Replying to @wrede
well I can't even open the latest cersion of their mobile site on my Windows phone (yeah, I know)
RT @snoaderDS: Love to read the praise for @zalandolounge and their unsubscribe process: https://t.co/84Icb65kq2 #ux #ecommerce #customerfi…
LinkedIn sure is the most wildly inconsistent product across devices and platforms I know and use.
1 6
Replying to @ayirpelle
@zientekglobal @robinwauters well, Zalando is certainly younger, but judging by numbers from wikipedia, we're much bigger now.
1
Replying to @ayirpelle
@zientekglobal @robinwauters it is! ;) You probably don't know about Zalando if we don't do business in your country...
1
Replying to @ayirpelle
@zientekglobal @robinwauters w00t that's a lot or w00t that's much less than I expected? ;)
1
RT @JennyBryan: A data scientist should be able to ... My favourite slide from @joelgrus's fun talk #ddtx16 https://t.co/GBlaR7Y5Yf
Media
If there is one thing the Berlin public transport busses are good at, it's heating. #oberdeck 😥
1 3
RT @swardley: This history of every NEW technology company ever ... https://t.co/TWIz9VGqRe
Media
RT @ApacheFlink: Anyone still using Hadoop 1? We are thinking about dropping Flink's support for Hadoop 1.
Replying to @huitseeker
Sorry, the edits were so complex it just HAD to take half a year to review it :'-|
Replying to @karpathy
should teach people is how that can be made fully composable with the chain rule.
4
Replying to @karpathy
I also searched extensively but didn't find anything good. Most textbooks do the simplest three Layer case. But what you really
1
Replying to @sszuecs
man, rails was good. I also think the Ruby community had a nice aporoach to building libraries sometimes.
1
Replying to @semithinking
oooh yeah! Of the most painful kind, because you're constantly aware that it's exactly that. ;)
@trieloff thanks ;) Only two more days, then we're supposed to be back in >0C regime again.
RT @ZalandoTech: Welcome back #Zalando #tech team! We are all looking forward to another ace year at #ZalandoTech! https://t.co/rDgPOeRjrU
RT @munterluggauer: @Atmel @mikiobraun and in the transition handbook the internet was still called catenet | justsaying
So we tried to stream Spotify from a Samsung phone to a Pioneer amp via WiFi Direct. Total protocol fail. This IS 2016, right?
1 2
RT @Atmel: #OnThisDay in 1983, the TCP/IP protocol went into effect on the ARPANET. https://t.co/dAhzhiH1Hi
Media
Replying to @bjornstenger
thankfully, I think it is mostly in my head ;) Well that and having been in academia for so long...