The course covers the foundations and modern methods of deep learning. Topics include multilayer neural networks, computational graphs and backpropagation, methods for stable training and generalization, convolutional networks and ResNet, representation learning, and self-supervised learning. RNN and LSTM models are studied as an introduction to attention mechanisms and Transformer architectures. The course also covers pretraining, fine-tuning, parameter-efficient adaptation methods such as LoRA, and modern generative models, including VAEs, autoregressive models, and diffusion models. Multimodal models are surveyed, and the course includes hands-on experience with PyTorch.
