IanGoodfellow PhD Defense Presentation

This video is the presentation of the PhD Defense of Ian Goodfellow.
There is the quick introduction in French, but the presentation is in English.
The slides: drive.google.com/file/d/0B640...
The thesis: drive.google.com/file/d/0B640...
Further information about each of these projects can be found at these links:
Spike and slab sparse coding (www-etud.iro.umontreal.ca/~goo...)
Multi-prediction deep Boltzmann machines (www-etud.iro.umontreal.ca/~goo..., techtalks.tv/talks/joint-train...)
Maxout networks (jmlr.org/proceedings/papers/v2..., techtalks.tv/talks/maxout-netw...)
Street View address number transcription (arxiv.org/pdf/1312.6082.pdf, • ICLR14: I Goodfellow: ... )

[@jokersmith9096]

No idea how the algorithm brought me here but I'm honoured.

[@nhatminhdao9480]

bro same here

[@gut_ton]

We both probably watch Lex fridman`s channel regularly

[@falling_water]

@@gut_ton not me.. Never heard of that name.

[@dayooladipo2992]

@@gut_ton Me too. Lex from MIT? Teaches machine learning

[@evilbabaroga]

watch the video and youll learn how

[@pfever]

Ian such a good fellow

[@thattimestampguy]

0:40 Thanks Jury 1:14 Fragmented presentation, with loose relations between articles 1:47 Solving Computer Vision Problems 2:18 The Ability of Software to learn from experience = Machine Learning 2:57 + Maximum Likelihood estimation/prediction (good predictions need strong parameters) 3:46 Gradient Descent 4:37 Supervised Learning (an input x, and a target output, y) 5:07 Unsupervised Learning 6:05 Deep Learning 7:55 Spike-and-Slab Sparse Coding. Motivating CIFAR-10 Results 9:14 Variational Learning ( A Dream, Wavy, Nasty to compute, hard to work with) 11:02 Analytical & Iterative Search Spike Variables, and Slab Variables, Multiplied together 12:44 _Flaw of Scaling_ 15:42 CIFAR-100 Results 16:09 Transfer Learning Challenge 17:46 DBM Training

[@omniyambot9876]

fuck u dude.

[@TheDeboAdedeji]

Keep doing what you're doing you beautiful human bean.

[@jamesang7861]

why is GAN not in his defense?

[@MuhammadFirdaus-uc1uy]

@@jamesang7861 because he invented it after he graduated

[@AnyFactor]

0:07 most aggressive shush in the history of shushes.

[@ultrasound1459]

French people be like 💀

[@nikitasmolin6232]

You're hired. Welcome to McDonalds.

[@maxsimes]

nah, his qualification is (barely) sufficient, but he needs at least two decades of experience. so no, rejected.

[@vodkacannon]

Hired? You're overqualified, your job will be getting unemployment checks. #RealLife

[@PauloConstantino167]

LOL

[@peterxdr]

Are you kidding me? This level of expertise is barely enough to become the floor-sweeper at McDonalds. If he wants the real job, he will require 50 years of experience in a completely unrelated field and at least 6 PhDs. Then, we will think about giving you a job.

[@oxyplik3542]

Bad joke

[@higormonteiro7636]

Maybe the most watched PhD thesis defense ever?

[@SouthernHerdsman]

Growth is a beautiful thing.

[@billykotsos4642]

That's a nice piece of A.I. history right here

[@hieu1814]

And a legend is born !

[@gusbakker]

Yoshua Bengio in the front :)

[@jayshah2714]

LEGEND

[@dutabesarkonohagakure647]

Sekelas Ian Goodfellow bikin disertasi kaya gini, antusias yg dateng, luar biasa

[@humza9460]

Now he's the director of machine learning at Apple

[@nikhilweee]

is the Q&A available as well?

[@wszjzhang]

guess it's usually not public for thesis defenses

[@Yustiks]

Great Work!

[@oxyplik3542]

8 years , this is not bad , iam here and it's good

[@supriyanaik316]

Can you suggest some link related to regression in Deep learning

[@MuhammadFirdaus-uc1uy]

I view this defense presentation to prepare my defense presentation today

[@thomas_testing]

Moving to that GPU was a real pro move.

[@junkwangpyo]

good stuffs to learn machine learning

[@dr.debajyotibose2928]

Good ole days of small PPT size !

[@jamesang7861]

Why is GAN not in his defense?

[@gachiemchiep]

What a legend!

[@thattimestampguy]

Machines are learning/being programed to learn

[@BryceChudomelka]

I present, The GANfather

[@rudaibaadnin7516]

Legend

[@TheBjjninja]

GAN-father

[@miguelangelquicenohincapie2768]

Jurgen Schmidhuber is the GAN-father

[@tdot33367]

the Deep Learning Algorithm has bought us here

[@rcrazy5841]

Every Ph.D student should watch this in their first year.

[@paperstars9078]

why?

[@gopimeister]

​@Nspnspker spot on. this looks like a pretty standard thesis defence and is also pretty much inapplicable for fields outside ML, or even just CS for that matter.

[@mayukhdifferent]

Deep Silence at 00:07

[@TileBitan]

You can't learn deep learning without your mouth being deeply shut

[@thecheekychinaman6713]

You mean Ian GodEmperor

[@jay._zer]

I was recommended this video so here's a like. Also the person below this is adopted.

[@janeluooo]

Me as a stats undergrad: Emm...I do understand something so that's cool

[@1968digga]

I thought they only give u 10-20 minutes to talk about ur thesis,then question u for the rest of the process.

[@cameronspence7500]

Yep, I defended mine yesterday, only allowed 20 minutes. I fairly rushed through my slides, in the end spoke for only 10 minutes.

[@socialogic9777]

@@cameronspence7500 Did that have any effect on the work that you wanted to portray or what they should have asked you.

[@cameronspence7500]

​@@socialogic9777 In hindsight I should have perhaps talked a bit slower... but in the end these slides are only a reminder of what the examiners should already have read in the thesis, where all the detail can be found. They are kind of a conversation-starter; the candidate is being judged on their thesis, not on their slides. But having a short amount of time was fine - forced me to focus on the key elements. In the end the defence only went on for 90 minutes which was a huge relief.

[@socialogic9777]

@@cameronspence7500 thankyou for detailed answer sir

[@sarathsivaprasad5712]

interesting

[@theawantikamishra]

Wondering how suddenly youtube want me to understand deep learning.

[@farahkanwal8192]

Any one there for econometrics?

[@asishswain1259]

whats that

[@xXTriforceXx]

This guy is obviously an absolute genius, but damn these slides are bland af.

[@theuberman7170]

I dont know about genius. This looks pretty standard for a PhD student

[@SkillazHDTV]

well what do you want? you want the slides to be dancing around?

[@MidStomp]

This is pretty standard concepts in intro machine learning now, pretty awesome

[@haleyharding4732]

@@MidStomp Yeah this guy just invented generative adversarial networks. Pretty standard tho hahaha give me a break

[@sirporkrib7643]

@@haleyharding4732 Isn't "his thesis is standard concepts in intro machine learning now" a huge compliment?

[@liamgoldbeck]

Why am I here?

[@davidzhang9388]

Why is the intro in French????

[@nouiz]

Because this is a French university.

[@hanifekri2612]

Don't forget that the university is locating in Canada.

[@jeromeblanchet3827]

Canada...Québec province...Montreal...best AI place ever...yeaaa, even better than OpenAI and stuff

[@TheNinjaDwarfBiker]

because it's in canada

[@MrGotickiller09]

Université de Montréal (University of Montreal) undergrad represent

[@Newtube_Channel]

Better not say anything short of being mean.

[@CriterionCafe]

Wow the intro was super Quebecois

[@Fnargl99]

Google algorithm

[@taha9892]

Is someone else sometimes sad how limited you are in understanding these kind of things? Maybe just me.

[@colaturkalures]

if you were dumber you wouldn't be sad

[@mahammad]

Me struggling to know what 10x10 is 😂😂

[@LuisGarcia1992_]

I can't judge the content, but the slides are REALLY ugly.

[@FellTheSky]

Who the fuck is ian and why did youtube brought me here

[@arefeshghi]

One of the most boring presentations are PhD completion ones! :) You have to sit there for an hour and pretend you are interested, just because your friend is presenting! :)) No KZbin, I won't watch this one! :))

[@mrborat2493]

It's interesting when you know who this guy is and what is he presenting

[@arefeshghi]

@@mrborat2493 I don't know this guy (however, I'm sure he is great), but I am a ML specialist and had my Bachelors, Masters and PhD in Computer Science. So, I know a thing or two about the subject. But still not wiling to sit through the whole thing! :)

[@AcceleratorUlz]

@@arefeshghi who cares

[@bright1402]

Legend