University of Toronto's CSC413: Deep Learning and Neural Network Course.
I took this course in 2021 Winter with prof Jimmy Ba and Bo Wang. There were 4 programming assignments, 4 math assignments, and a final project(also a midterm). It introduced a broad range of deep learning topics(CNN, RNN, GAN etc.) and moreover, some state-of-the-art models (BERT, Style2-GAN etc). As a fourth-year course, it was quite demanding, yet very interesting and made me learn so much. I really recommend it.
We evaluated two text classification methodologies, Simple Graph Convolution (SGC) and Neural Attentive Bag-of-Entities(NABoE), by performing hyperparameter sensitivity analysis, extending to a new dataset, and experimenting with the model architechtures. The final report was written in NeurIPS format.
PA1: Word Embedding with GloVE, Masked Language Modelling
PA2: Convolutional Neural Networks(CNN) for Image Classification
PA3: Attention-Based Neural Machine Translation: LSTM, Additive Attention, Scaled Dot Product Attention
PA4: Deep Convolutional GAN (DCGAN), StyleGAN2-Ada, Deep Q-Learning Network (DQN)
MA1: Hard-Coding Networks, BackPropagation, Linear Regression
MA2: Optimization, Gradient-based Hyper-parameter Optimization, Convolutional Neural Networks
MA3: Robustness and Regularization, Trading off Resources in Neural Network Training, Dropout as Gaussian Noise
MA4: Recurrent Neural Network(RNN) and Self-attention, Reversible Models and Variational AutoEncoders(VAE), Reinforcement Learning
https://csc413-uoft.github.io/2021/ (May be deactivated in the future)