Thesis Description & Objectives
Neuro-symbolic AI encompasses a broad class of architectures aiming at combining machine learning with knowledge representation and reasoning (“Good old-fashioned AI”). In particular, Logic Tensor Networks are a framework to encode a first order logic language into a trainable neural network. NeSy techniques like LTNs introduce many additional design choices, such as: How to define axiomatic prior knowledge (“how to encode the rules of Sudoku”?); How to ground predicates and connectives? How to choose between different frameworks? Comparison about existing frameworks is limited in literature, and issues in reporting, hyper-parameter selections, etc. affect this comparison.
Several thesis are available on tackling these issues. Depending on the candidate background and predisposition, the problem can be tackled either from a more theoretical or experimental standpoint. Starting from an analysis of the current literature, techniques to be investigated include:
– Analysis of the mathematical and numerical properties of different frameworks
– Comparison of existing frameworks and losses
– Application of techniques from the fields of hyper-parameter optimization (Random search, Bayesian Optimization), Neural Architectural Search (NAS), genetic algorithms and curriculum learning
Experiments will be conducted on standard benchmarks as well as applications on semantic image interpretation explored by the group.
Prerequisites: programming skills (Python, Pytorch or other deep learning framework); good analytical skills. Prior knowledge of neuro-symbolic techniques is not required – essential material to study on the topic will be provided. The candidate we are looking for is highly motivated and interested towards research-oriented activities.
