HKUST TensorFlow 教程笔记。

Day 1

  • TensorFlow installation
    • pip install (GPU version only for Linux)
    • Build from source
  • TensorFlow basic usage
    • Check installation and version
    • Hello world
  • TF Mechanics
    • Computational Graph
    • Placeholder
    • Tensor: concept, ranks, shapes and types
  • Machine learning basics
    • Linear Regression
      • Gradient descent algorithm
    • Logistic Regression
      • Sigmoid function
    • Softmax Classification
      • Cost: cross entropy
      • epoch, batch size, iterations

Day 1.5

  • XOR with Logistic Regression
    • Single unit/Multiple units
  • Solutions: Deep Neural Network
  • Challenges
    • Forward propagation/Back propagation(Chain rule)
    • Activation units: ReLU
    • Overfitting
    • Regularization/Dropout
  • TensorBoard

Day 2

  • Convolutional Neural Network
    • width, height, depth
    • filter
    • Convolutional Layer
      • CONV + ReLU
    • Pooling Layer
      • downsampling (Max pooling)

Day 3

  • RNN (left for future)

留言