| KEYWORD |
DAUIN - GR-09 - GRAphics and INtelligent Systems - GRAINS
Solving Sudoku puzzles with neuro-symbolic techniques
keywords ARTIFICIAL INTELLIGENCE, DEEP LEARNING, KNOWLEDGE REPRESENTATION AND REASONING
Reference persons LIA MORRA
Research Groups DAUIN - GR-09 - GRAphics and INtelligent Systems - GRAINS
Description While state-of-the-art deep learning typically uses distributed representations, reasoning is normally useful at a higher level of abstraction. As a result, attempts at combining symbolic AI and neural computation into neural-symbolic systems have been on the increase. Neural-symbolic learning and reasoning is an area of research that aims at integrating relational/symbolic reasoning and knowledge representation (e.g., first order logic language for knowledge representation) by embedding their constructs in deep neural networks. NeSy techniques can be used to solve tasks such as Visual Sudoku Puzzle classification, whose goal is to automatically check the validity of handwritten solutions for Sudoku. The system must learn to distinguish digit by satisfying the rules of Sudoku, without providing any digit label. However, 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.
The goal of this thesis is to experimentally and extensively investigate the effect of different design choices on the task of visual Sudoku puzzle classification, from the formulation of the constraints to the hyper-parameter settings, on Logic Tensor Networks designed for this task. In addition, LTNs will be compared against other neuro-symbolic frameworks (such as NeuPSL) and standard deep learning baselines (such as visual transformers and relational networks).
Reading list:
Morra, Lia, et al. "Designing Logic Tensor Networks for Visual Sudoku puzzle classification." 17th International Workshop on Neural-Symbolic Learning and Reasoning (NeSy 2023). CEUR, 2023.
https://www.cs.ox.ac.uk/isg/conferences/tmp-proceedings/NeSy2023/paper19.pdf
Deadline 28/02/2025
PROPONI LA TUA CANDIDATURA