Gradient-based Learning of Complex Latent Structures

Imposing structural constraints on the latent representations learned by deep neural models has several applications, which can improve their explainability, generalisation, and robustness properties. For example, we can learn more explainable models by making them selectively decide which parts of the input to consider; or we can improve their generalisation properties by learning representations suitable for reasoning tasks, such as deductive reasoning and planning, and comply with any desired constraints. The main reason why learning complex discrete – or mixed continuous-discrete – representations is not widely popular is that back-propagating discrete decision steps used to be problematic, but several practical solutions have been proposed recently (e.g., see Niepert et al., 2021; Minervini et al., 2022). In this project, we investigate how we can derive better methods for back-propagating through mixed continuous-discrete complex latent structures and how we can leverage them for learning more explainable, data-efficient, and robust deep neural models.