The "Deep Learning" course focuses on advanced deep learning technologies and architectures that shape the modern landscape of artificial intelligence. It aims at understanding theoretical foundations, practical implementation of advanced models, and ethical development of artificial intelligence systems.
The course content includes fundamental principles such as computation graphs, activation functions, optimization algorithms, and regularization techniques. It covers classical deep neural network architectures (CNNs, RNNs, etc.), architectures for graph-structured data (GNNs), as well as modern architectures such as Transformers, Vision Transformers, and Foundation models.
Additionally, self-supervised learning and contrastive learning techniques for representations from unlabeled data are introduced, as well as multimodal learning techniques. The course also includes generative models such as autoencoders (AEs, VAEs), generative adversarial networks (GANs), and diffusion models. Tensor methods for model interpretation and optimization are examined. Finally, ethical challenges and ethics issues in developing artificial intelligence systems are discussed.