Elena Umili, Francesco Argenziano, Roberto Capobianco
DeepDFA is a framework that integrates temporal logic into neural networks, outperforming traditional models in sequential tasks by bridging subsymbolic learning and symbolic reasoning.
DeepDFA is a new approach that combines logical rules with deep learning to improve how machines understand sequences of data, like series of images or steps in a process. By using something called Deterministic Finite Automata, DeepDFA can incorporate high-level logical rules directly into neural networks. This helps the networks perform better in tasks that involve understanding sequences, such as classifying image sequences or learning policies in environments that don't follow simple patterns. In tests, DeepDFA showed better performance compared to other models, suggesting it could be a powerful tool for tasks that require both learning from data and applying logical rules.
