Complete Statistical Theory of Learning (Vladimir Vapnik) | MIT Deep Learning Series

Lecture by Vladimir Vapnik in January 2020, part of the MIT Deep Learning Lecture Series.

OUTLINE:
0:00 - Introduction
0:46 - Overview: Complete Statistical Theory of Learning
3:47 - Part 1: VC Theory of Generalization
11:04 - Part 2: Target Functional for Minimization
27:13 - Part 3: Selection of Admissible Set of Functions
37:26 - Part 4: Complete Solution in Reproducing Kernel Hilbert Space (RKHS)
53:16 - Part 5: LUSI Approach in Neural Networks
59:28 - Part 6: Examples of Predicates
1:10:39 - Conclusion
1:16:10 - Q&A: Overfitting
1:17:18 - Q&A: Language

CONNECT:
- If you enjoyed this video, please subscribe to this channel.

[lexfridman]

I really enjoyed this talk by Vladimir. Here's the outline:
0:00 - Introduction
0:46 - Overview: Complete Statistical Theory of Learning
3:47 - Part 1: VC Theory of Generalization
11:04 - Part 2: Target Functional for Minimization
27:13 - Part 3: Selection of Admissible Set of Functions
37:26 - Part 4: Complete Solution in Reproducing Kernel Hilbert Space (RKHS)
53:16 - Part 5: LUSI Approach in Neural Networks
59:28 - Part 6: Examples of Predicates
1:10:39 - Conclusion
1:16:10 - Q&A: Overfitting
1:17:18 - Q&A: Language

[ephi]

My master's thesis was forecasting using SVM. That was the first time I fell in love with machine learning and even Math. Thank you Vladimir for living.

[rodolfo_bandeira]

I'm studying SVM in my MCS program. I was so surprised to find this video with Dr. Vapnik. We live in such blessed times to have easy access to this level of high-quality content.
Thank you!

[alchemication]

Huge respect for the gentleman (he is a legend for us, AI-Masters students in Ireland;) Thank you for uploading to YouTube.

[SuperReminou]

I feel privileged to have the opportunity to watch this video. Thank you very much @Lex Fridman

[Anza_]

His concept of predicates is intriguing: Everything can be deconstructed to see what it is consisting of - the basic building blocks. With that, what is left to do is only one more step: analyzing the structure.
Excellent concept!

[rezab]

@Lex Fridman 1.5 years ago I listened to your first podcast with prof vapnik and was blown away. Great man, great story. I love it. Funny is that while pursuing the topic of machine learning and deep learning myself at the moment I hit the subject of learning curves, cross-validation and other methods to learn more efficient and remembered the podcast in which he mentioned his Complete Statistical Theory and as a former math major I appreciate his approach so much. Thx for this opportunity

[butterkaffee]

This envokes great memories to my university days! Working in applied ml is seldom as elegant as vc theory lol

[TheAIEpiphany]

I think this lecture broke my mind. Legend!

[davidbellamy]

Lex, we are so grateful for the amazing lectures and conversations you provide to the Internet all assembled in one place, thank you!

[StratosFair]

Wow, being taught by the man himself, what an honor

[prattzencodes]

It's an honor to see one of the living legends of Theoretical Machine Learning / and the father Statistical Learning Theory in flesh! <3

[DiegoAToala]

Thank you very much for this video. Watching a lecture from this gentleman is such an honor.

[evankim]

Amazing talk and amazing contributions to the field of statistical learning theory. This is definitely a piece of the puzzle that I feel like is very under represented today.

[harryh]

Funnily, YouTube detects the language of the Video as Russian for subtitles

[alexanderkonstantinidis]

Thank you for offering us this possibility.

[oudarjyasensarma]

Thanks a lot!!! Very Informative!!!! And thanks for making all of this happen!!!!

[madtrade]

i use his invention every single day !

[oudarjyasensarma]

HI Lex :), Can you do a series/playlist on NLP research and where NLP is going after 2020 and its future? That'd be really helpful!!!!

[priapushk]

0:04 "co-inventor of supported vector machines". Lex invented the unsupported ones.