Chapter 1. Neural-Symbolic Learning and Reasoning: A Survey and Interpretation1

Abstract

The study and understanding of human behaviour is relevant to computer science, artificial intelligence, neural computation, cognitive science, philosophy, psychology, and several other areas. Presupposing cognition as basis of behaviour, among the most prominent tools in the modelling of behaviour are computational-logic systems, connectionist models of cognition, and models of uncertainty. Recent studies in cognitive science, artificial intelligence, and psychology have produced a number of cognitive models of reasoning, learning, and language that are underpinned by computation. In addition, efforts in computer science research have led to the development of cognitive computational systems integrating machine learning and automated reasoning. Such systems have shown promise in a range of applications, including computational biology, fault diagnosis, training and assessment in simulators, and software verification. This joint survey reviews the personal ideas and views of several researchers on neural-symbolic learning and reasoning. The article is organised in three parts: Firstly, we frame the scope and goals of neural-symbolic computation and have a look at the theoretical foundations. We then proceed to describe the realisations of neural-symbolic computation, systems, and applications. Finally we present the challenges facing the area and avenues for further research.

Discrete and continuous representations and processing in deep learning: Looking forward

“…Finally, we note that disagreements between symbolic AI and connectionism and statistical learning in general have occurred recurrently throughout the history of AI, as shown by Minsky and Papert (1969); Marcus (2018). Alongside those tensions, there have also been calls for the combination of elements of both fields (Fodor et al, 1988;Minsky, 1991;Besold et al, 2017;Lake et al, 2017;Marcus, 2020), and correspondingly, there have already been many attempts to do so (Sun and Bookman, 1994;d'Avila Garcez et al, 2009;Manhaeve et al, 2018;Yi et al, 2018). However, the message that we want to convey with this discussion, as summarized by the last sentences of section 4 and the first sentences of this section, is more general than a call for the integration of classical AI with connectionism.…”

Faster-LTN: a neuro-symbolic, end-to-end object detection architecture

“…Several techniques have been proposed to augment Convolutional Neural Networks (CNNs) with relationship representation and reasoning capabilities, including Relational Network [6], Graph Neural Networks [7] and Neural-Symbolic (NeSy) techniques [8,3,5]. For a more general introduction to NeSy techniques, the reader is referred to recent surveys [9,10].…”

Generative Flow Networks for Discrete Probabilistic Modeling

“…Probabilistic modeling in discrete spaces, especially those with compositional structure, is important due to the universality of applications of discrete data structures, such as in natural language processing (Tai et al, 2015) or in symbolic reasoning (Besold et al, 2017). However, distributions in high-dimensional discrete spaces are generally hard to model, as they may feature rapid combinatorial growth of modes.…”