Deep Learning and Generative AI (AOT509)

This course covers core deep learning techniques and the evolution of generative models. Topics include neural networks, CNNs, RNNs, autoencoders, GANs, diffusion models, and the Transformer architecture. Students will also explore large language models, prompting, fine-tuning, and ethical considerations, preparing them to apply and evaluate cutting-edge generative AI technologies.

• Introduction to Neural Networks and Machine Learning
• Training Neural Networks: Backpropagation and Gradient Descent
• Convolutional Neural Networks (CNNs) for Computer Vision
• Recurrent Neural Networks (RNNs) for Sequential Data
• The Generative Idea: Autoregressive Models and Autoencoders
• Variational Autoencoders (VAEs) for Latent Space Generation
• Generative Adversarial Networks (GANs)
• Advanced GANs and Introduction to Diffusion Models
• The Transformer Architecture and the Attention Mechanism
• The Rise of Large Language Models (LLMs)
• Prompting, Fine-Tuning, and In-Context Learning
• Multimodality, Ethics, and the Future of Generative AI

• Assess the ethical, societal, and technical implications of generative AI, particularly in multimodal and autonomous systems.
• Evaluate the strengths and limitations of generative models including autoencoders, VAEs, GANs, and diffusion models across various applications.
• Formulate training strategies using backpropagation, gradient descent, and hyperparameter tuning to optimise neural network performance.
• Set up deep learning architectures such as CNNs, RNNs, and Transformers to solve complex tasks in vision, language, and sequential data.
• Recommend optimised large language models (LLMs) for specific tasks using prompting and in-context learning techniques.
• Design state-of-the-art deep learning and generative AI concepts to create innovative solutions for real-world problems.