The latest advances in machine learning (ML) can be mainly attributed to deep learning (DL); more specifically, to representation learning (i.e., deep neural networks extracting meaningful patterns from raw data to create representations that are easier to understand and process). In particular, recent years have seen the success of such representation learning techniques in the form of generative AI, in domains such as vision, language and multi-modal data. Therefore, OxML2024 will dedicate a module to this topic; we will bring together some of the world’s top researchers (academia and industry) in representation learning and generative AI (both theory and applied), to cover the latest developments and the current state-of-the-art in areas such as:

​

• Advanced topics in representation learning and computer vision

• Advanced topics in representation learning and sequences (e.g., language models)

• Advanced topics in multi-modal representation learning

• The current state of foundational models in vision, language, …

• Geometrical deep learning

• Reinforcement learning 

• Innovations in training neural networks (e.g., contrastive learning, self-supervised learning, …)

• New (and improved older) neural architectures (e.g., hopfield networks)

• Knowledge graphs, knowledge-aware ML, and neuro-symbolic ML

• Taking highly-capable neural networks to real-world products (e.g., RLHF, alignment, …)