課程概述 |
深度學習是人工智慧領域中近年發展最為迅速的領域。深度學習最近在影像辨認、電腦視覺、語音辨認、資訊安全、自然語言處理、電腦遊戲等眾多領域均以優異效能領先其他。本課程內容包含深度學習模型、學習理論、並透過實作基本最佳化理論的梯度下降於不同成本函數(MSE, Cross-Entropy),了解反向傳播學習理論,並進而從開源系統專案,實際運用深度學習模型(MLP,CNN, ResNet, LSTM, GRU)於大型資料集(Fashion MNIST, CIFAR10, CIFAR100)了解深度學習技術。
課程大網包括以下項目:
1. Introduction to Deep learning, Environment Installation
2. Perceptron、Linear model, Basic optimization
3. Logistic Regression, Softmax model, Cross-entroy v.s. MSE
4. Multilayer Perceptron, Universal Function Approximation Theorem
5. Matrix Calculus and Back propagation algorithm
6. Advanced Optimization, Momentum
7. Weight Initialization, Batch Normalization, Weight Decay
8. The difficulty of training deep networks: vanishing/exploding gradients
9. Convolution Neural Network, Image Classification
10. Object Detection, Yolo, SSD
11. Image Style Transfer
12. LeNet, AlexNet, VGG, Inception, Residual Network
13. Recurrent neural network, Gated Recurrent Unit, Long-Short Term Memory
14. Deep Generative Models: Autoencoder, Generative Adversarial Networks
|
參考書目 |
參考教材:
(1) Deep Learning, Ian Goodfellow, Yoshua Bengio and Aaron Courville
(2) A Programmer's Guide to Data Mining: The Ancient Art of the Numerati, Ron Zacharski
(3) Python Data Science Handbook: Essential Tools for Working with Data, Jake VanderPlas , O'Reilly Media; 1 edition (December 10, 2016).
|