Anchoring Knowledge in Interaction: Towards a Harmonic Subsymbolic/Symbolic Framework and Architecture of Computational Cognition

Besold, Tarek R.; Kühnberger, Kai‐Uwe; Garcez, Artur S. d’Avila; Saffiotti, Alessandro; Fischer, Martin H.; Bundy, Alan

doi:10.1007/978-3-319-21365-1_4

Cited by 3 publications

(3 citation statements)

References 13 publications

(16 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Against this background, in Besold et al (2015) a program has been proposed 530 for 'anchoring knowledge in interaction', aiming at developing, theoretically and practically, a conceptual framework and corresponding architecture that model an agent's knowledge, thinking, and acting truly as interrelated parts of a unified cognitive capacity. That is, knowledge is seen as multi-layered phenomenon that appears at different levels of abstraction, promotes interaction between 535 these levels of abstraction, is influenced by the interaction between agent and environment (potentially including other agents), and is essentially linked to actions, perception, thinking, and being.…”

Section: From Interaction To Knowledge and Backmentioning

confidence: 99%

Towards integrated neural–symbolic systems for human-level AI: Two research programs helping to bridge the gaps

Besold

Kühnberger

2015

Biologically Inspired Cognitive Architectures

Self Cite

View full text Add to dashboard Cite

This is the accepted version of the paper.This version of the publication may differ from the final published version. signals, but cognition also exhibits the capability to efficiently perform abstract reasoning and symbol processing; in fact, processes of the latter type seem to form the conceptual cornerstones for thinking, decision-making, and other (also directly behavior-relevant) mental activities (see, e.g., Fodor & Pylyshyn (1988)). (2015)). This has several reasons also beyond the 35 analogy to the described functioning principles of the brain: Permanent repository link• In general, network-based approaches possess a higher degree of biological motivation than symbol-based approaches, also outmatching the latter in

show abstract

Section: From Interaction To Knowledge and Backmentioning

confidence: 99%

Towards integrated neural–symbolic systems for human-level AI: Two research programs helping to bridge the gaps

Besold

Kühnberger

2015

Biologically Inspired Cognitive Architectures

Self Cite

View full text Add to dashboard Cite

show abstract

“…One of the biggest challenges in this context is the (bidirectional) integration between low-level sensing and interacting as well as the formation of high-level knowledge and subsequent reasoning. In (Besold, Kühnberger, Garcez, Saffiotti, Fischer, & Bundy, 2015) an interdisciplinary group of researchers spanning from cognitive psychology through robotics to formal ontology repair and reasoning sketch conceptually (what amounts to) a strongly cognitively-motivated neural-symbolic architecture and model of a situated agent's knowledge acquisition through interaction with the environment in a permanent cycle of learning through experience, higher-order deliberation, and theory formation and revision.…”

Section: Other Research Efforts and Programs Narrowing The Neural-sym...mentioning

confidence: 99%

Neural-Symbolic Learning and Reasoning: A Survey and Interpretation

Besold,

Garcez,

Bader

et al. 2017

Preprint

Self Cite

View full text Add to dashboard Cite

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 1

show abstract

“…In this paper, the distinct components of this challenge are isolated and investigated systematically for the first time, towards a better understanding of the fundamental memory properties of the current state-of-the-art language models based on recurrent neural networks (RNNs). This paper proposes five basic tasks for isolating and examining specific capabilities relating to the implementation of memory in RNNs, and analyses the ability of such networks to solve these problems in order to gain insight into the fundamental intra-cell and inter-cell mechanisms that RNNs learn to employ, thus contributing to the research on the use of RNNs, and focusing particularly on work in the last two years addressing the ability of RNNs to implement memory and perform symbol grounding as well as reasoning [3], [4], [5]. Specifically, the long short term memory (LSTM) cell [3], originally developed to address the vanishing gradient problem, is evaluated in comparison with the gated recurrent unit (GRU) [4].…”

Section: Introductionmentioning

confidence: 99%

On the memory properties of recurrent neural models

Russell

Benetos

Garcez

2017

2017 International Joint Conference on Neural Networks (IJCNN)

View full text Add to dashboard Cite

In this paper, we investigate the memory properties of two popular gated units: long short term memory (LSTM) and gated recurrent units (GRU), which have been used in recurrent neural networks (RNN) to achieve state-of-the-art performance on several machine learning tasks. We propose five basic tasks for isolating and examining specific capabilities relating to the implementation of memory. Results show that (i) both types of gated unit perform less reliably than standard RNN units on tasks testing fixed delay recall, (ii) the reliability of stochastic gradient descent decreases as network complexity increases, and (iii) gated units are found to perform better than standard RNNs on tasks that require values to be stored in memory and updated conditionally upon input to the network. Task performance is found to be surprisingly independent of network depth (number of layers) and connection architecture. Finally, visualisations of the solutions found by these networks are presented and explored, exposing for the first time how logic operations are implemented by individual gated cells and small groups of these cells.

show abstract

Anchoring Knowledge in Interaction: Towards a Harmonic Subsymbolic/Symbolic Framework and Architecture of Computational Cognition

Abstract: This is the accepted version of the paper.This version of the publication may differ from the final published version. Permanent repository link:

Cited by 3 publications

References 13 publications

Towards integrated neural–symbolic systems for human-level AI: Two research programs helping to bridge the gaps

Towards integrated neural–symbolic systems for human-level AI: Two research programs helping to bridge the gaps

Neural-Symbolic Learning and Reasoning: A Survey and Interpretation

On the memory properties of recurrent neural models

Contact Info

Product

Resources

About