Human reasoning seminars: Simon Dobnik, Logic in neural networks

Seminarium

Simon Dobnik från King¿s College London kommer att ge ett seminarium om: Logik i neurala nätwerk (på engelska).

Simon Dobnik

Logic in neural networks

Abstract. I will present a paper by Garcez and Lamb (2011) that discusses recent advances in representing logics (with emphasis on modal and temporal logic) in artificial neural networks (ANNs). Logics have clear semantics, automated proof methods, and proof histories which provide an explanation of the reasoning process. On the other hand ANNs are very efficient and robust in learning from examples by capturing the statistical properties of a dataset. Applying the models that have been learned in classification provides deduction. The joint properties of both are very attractive for representating and reasoning about language and perception which has been my interest. I will conclude by sketching some examples. Garcez, Artur S. d'Avila and Lamb, Luis C. 2011. Cognitive Algorithms and Systems: Reasoning and Knowledge Representation. In Perception-Action Cycle, edited by Vassilis Cutsuridis, Amir Hussain, and John G. Taylor. Springer Series in Cognitive and Neural Systems. 573-600. http://dx.doi.org/10.1007/978-1-4419-1452-1_18