Deep Learning Short Course - 2018

Course Description

This is a short course on deep learning. It introduces general concepts in AI, machine learning, and deep learning. The course reviews popular architectures in deep neural networks and techniques frequently used by practitioners.

Schedule

  Lecture Lab
1
  • Introduction: Part 1
        - AI
        - Traditional AI
        - Machine learning
        - Generalization error
        - Three major ML problems
        - Linear regression
  • TensorFlow
  • Linear Regression
2
  • Introduction: Part 2
        - Linear classification
        - Artificial neural networks
        - Deep learning
        - Deep reinforcement learning
        - Deep learning: some recent applications
  •  MNIST Classification Using a Multilayer Perceptron (MLP)
3
  • Convolutional Neural Networks (CNN) [Part 1, Part 2]
        - Convolution
        - ReLU, Pooling
        - Stochastic gradient descent: Momemtum, RMSprop, ADAM
        - Dropout
        - Batch normalization
        - Popular CNN architectures: AlexNet, VGGNet, ResNet
  •  MNIST Classification Using CNNs
4
  • Recurrent Neural Networks (RNN)
        - RNNs
        - Long short-term memory (LSTM)
        - Gated recurrent unit (GRU)
        - Sequence-to-sequence (Seq2Seq) model
  •  CIFAR-10 Image Classification
5
  • Advanced Topics
        - Generative adversarial networks (GAN)
        - Nested sparse networks
        - Deep reinforcement learning: MDP, DQN, sparse RL
  •  Name Generation Using RNNs