STAT479: Deep Learning (Spring 2019)
Instructor: Sebastian Raschka
Lecture material for the STAT 479 Deep Learning course at University Wisconsin-Madison. For details, please see the course website at http://pages.stat.wisc.edu/~sraschka/teaching/stat479-ss2019/
Course Calendar
Please see http://pages.stat.wisc.edu/~sraschka/teaching/stat479-ss2019/#calendar.
Topic Outline
- History of neural networks and what makes deep learning different from “classic machine learning”
- Introduction to the concept of neural networks by connecting it to familiar concepts such as logistic regression and multinomial logistic regression (which can be seen as special cases: single-layer neural nets)
- Modeling and deriving non-convex loss function through computation graphs
- Introduction to automatic differentiation and PyTorch for efficient data manipulation using GPUs
- Convolutional neural networks for image analysis
- 1D convolutions for sequence analysis
- Sequence analysis with recurrent neural networks
- Generative models to sample from input distributions
- Autoencoders
- Variational autoencoders
- Generative Adversarial Networks
Material
- L01: What are Machine Learning and Deep Learning? An Overview. [Slides]
- L02: A Brief Summary of the History of Neural Networks and Deep Learning. [Slides]
- L03: The Perceptron. [Slides] [Code]
- L04: Linear Algebra for Deep Learning. [Slides]
- L05: Fitting Neurons with Gradient Descent. [Slides] [Code]
- L06: Automatic Differentiation with PyTorch. [Slides] [Code]
- L07: Cloud Computing. [Slides]
- L08: Logistic Regression and Multi-class Classification [Slides] [Code]
- L09: Multilayer Perceptrons [Slides] [Code]
- L10: Regularization [Slides] [Code]
- ...