Project participants collaborate on research about integrating machine learning and symbolic reasoning using neural networks. Neural-symbolic computation is an interdisciplinary research area borrowing from computer science, artificial intelligence, neural computation, machine learning, computational logic, cognitive and neurosciences, psychology and philosophy.
The goal of this project is to come to a unified architecture that supports symbolic learning and reasoning using neural networks.
Topics of interest include: Network reasoning and inference, Reasoning about uncertainty, Learning from structured data, Statistical relational learning, Integrated reasoning and learning, Expressive reasoning with robust learning, Nonclassical models of computation, Computational theories of mind, Cognitive computation, abduction and analogy, Knowledge extraction from complex networks, Deep learning and reasoning, Combination of systems, Efficient implementations of integrated learning and reasoning.
Applications in large-scale data analysis problems including simulation and training, robotics, the web, multi-agent systems, fault diagnosis, bioinformatics, argumentation, normative systems, security, multi-modal learning, visual information processing, anomaly detection, fraud prevention.