full stack deep learning
play

Full Stack Deep Learning Lecture 1: Intro Pieter Abbeel, Sergey - PowerPoint PPT Presentation

Full Stack Deep Learning Lecture 1: Intro Pieter Abbeel, Sergey Karayev, Josh Tobin Organizer Intros Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 1 Guest Instructors Saturday Sunday Sunday


  1. Full Stack Deep Learning Lecture 1: Intro Pieter Abbeel, Sergey Karayev, Josh Tobin

  2. Organizer Intros Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 1

  3. Guest Instructors Saturday Sunday Sunday Saturday Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 2

  4. Thanks to our sponsors! Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 3

  5. Why deep learning? Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 4

  6. Why is AI hard to build? E.g. vision? “C “Coffe ffee Mug” Pi Pixel el Inten ensi sity Pi Pixel el inten ensi sity y is s a a ver very y po poor repr epresen esentat ation. [slide from Adam Coates] Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 5

  7. Object Detection in Computer Vision • State-of-the-art object detection until 2012: Support Hand-engineered “cat” Input Vector features (SIFT, “dog” Image Machine HOG, DAISY, …) “car” … (SVM) • Deep Supervised Learning (Krizhevsky, Sutskever, Hinton 2012; also LeCun, Bengio, Ng, Darrell, …): “cat” Input 8-layer neural network with 60 million “dog” Image parameters to learn “car” … • ~1.2 million training images from ImageNet [Deng, Dong, Socher, Li, Li, Fei-Fei, 2009] Pieter Abbeel -- Sergey Karayev -- Josh Tobin

  8. Neural Net Learning Image Recognition • Training / Learning: • Cars: labeled • Cats: data • Dogs: Machine Learning * Test Time: Label = Cat? Dog? Car? à à Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 7

  9. Many Layer Neural Network cat car dog nothing different weights à different computation Neural Net Training: Find the weights that minimize the difference between labels and activation. Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 8

  10. Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 9

  11. Performance graph credit Matt Zeiler, Clarifai Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 10

  12. Performance graph credit Matt Zeiler, Clarifai Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 11

  13. Performance graph credit Matt Zeiler, Clarifai Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 12

  14. MS COCO Image Captioning Challenge Karpathy & Fei-Fei, 2015; Donahue et al., 2015; Xu et al, 2015; many more Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 13

  15. Visual QA Challenge Stanislaw Antol, Aishwarya Agrawal, Jiasen Lu, Margaret Mitchell, Dhruv Batra, C. Lawrence Zitnick, Devi Parikh Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 14

  16. Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 15

  17. Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 16

  18. https://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm604357.htm Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 17

  19. Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 18

  20. Speech Recognition graph credit Matt Zeiler, Clarifai Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 19

  21. Machine Translation Google Neural Machine Translation (in production) Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 20

  22. History (Olshausen, 1996) 2000s Sparse, Probabilistic, and Energy models (Hinton, Bengio, LeCun, Ng) Rosenblatt’s Perceptron based on history by K. Cho Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 21

  23. Why Now? • Data • Compute • Some new ideas Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 22

  24. [Source: domo.com] Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 23

  25. Compute [Source: OpenAI] Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 24

  26. Why this class? Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 25

  27. Why this class? Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 26

  28. Why this class? n Bring people together to talk about the challenges of working on ML projects n We want to help those who want to work in the field of deep learning get their dream job Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 27

  29. How we developed content n We talked to deep learning engineers and leaders at Airspace, Symbio, Determined.ai, W&B, Forethought.ai, DeepScale, Google, Ancestry, Lambda Labs, Bloomberg Beta, Rent the Runway, Magic Leap, Facebook, Tesla, Uber, OpenAI, and more. Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 28

  30. Schedule Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 29

  31. Logistics Make sure you’re in our Slack workspace (email team@fullstackdeeplearning.com if you’re not) Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 30

  32. Questions? Full Stack Deep Learning (March 2019) Pieter Abbeel, Sergey Karayev, Josh Tobin L1: Intro 31

Recommend


More recommend