Implementation Aspects of Tensor Product Variable Binding in Connectionist Systems

Demidovskij A.

Tensor Product Variable Binding is an important aspect of building the bridge between the connectionist approach and the symbolic paradigm. It can be used to represent recursive structures in the tensor form that is an acceptable form for neural networks that are highly distributed in nature and, therefore, promise computational benefits from using it. However, practical aspects of tensor binding implementation using modern neural frameworks are not covered in the public research. In this work, we have made an attempt of building the topology that can perform binding operation for a well-known framework Keras. Also we make the analysis of the proposed solution in terms of its applicability for other important connectionist aspects of Tensor Product Variabl Binding. Proposed design of the binding network is the first step towards expressing any symbolic structure and operation in the neural form. This will make it possible for traditional decision making algorithms to be replaced with a neural network that brings scalability, robustness and guaranteed performance.

Keywords: artificial neural networks, Decision support systems, Tensor computations

In book

Advances in Intelligent Systems and Computing

Vol. 1037: Proceedings of the 2019 Intelligent Systems Conference (IntelliSys). , Switzerland: Springer, 2019.