Deep Learning - What is it and What It Can Do For You
summary
Explains how deep learning composes differentiable modules with a forward pass and a backward pass implemented via the chain rule and reverse-mode automatic differentiation so you can train parameters by gradient descent, and reviews concrete modules: matrix multiplies, ReLU/softmax, convolutions, RNNs and LSTMs. Claims include the ImageNet 2012 jump that transformed computer vision, medical applications like lung cancer detection from CTs, Atari and AlphaGo game results, robots learning grasping and Nvidia driving from a single camera. Mentions tools such as TensorFlow and Theano and the CS231n course.
AI-written summary · about
chapters (25)
- 0:00Introduction
- 1:25Super High level
- 2:52Deep Learning An Interface
- 4:06Deep Learning Vocabulary
- 5:52Details: How to implement the backward pass
- 7:12Details: How do we update the parameters
- 7:32Details: What is backproplagation
- 8:40Basics Review
- 9:38Matrix Multiplication
- 10:28Sigmoid
- 11:38Slight Tangent
- 17:47Recurrent Neural Networks CRNNO
- 23:07Convolutional Neural Networks (CNN)
- 36:38Robot Overlords
- 38:12Generation
- 40:38Deeper Language Understanding
- 41:34Deeper Visual Understanding
- 42:06Harder Game Playing
- 42:54Automatically Discovering Hierarchical Structure
- 43:47Learning to Use Memory
- 44:06Differentiable Data Structures
- 44:33Learning to Program
- 45:02Learning to Execute
- 45:37Module Networks
- 46:29OKI I'm pumped to deep learning some problems
related talks
54mIntroduction to Understanding Deep Learning
27mDemystifying Deep Neural Networks
45mIntroduction To Deep Learning
35mDeconstructing Deep Learning
01h 37mDeep learning tutorial - advanced techniques
01h 04mWhat you always wanted to know about Deep Learning, but were afraid to ask
37mWhat's new in Deep Learning?
32mDeep Learning, What Are You Good For!