Conceptual model of a cognitive system: dynamical system with plastic self-organizing velocity field

TL;DR

This paper introduces a new class of dynamical systems called self-shaping DSs with plastic velocity fields, modeling cognitive systems capable of online, unsupervised learning without spurious attractors, demonstrated through a musical example.

Contribution

It presents a novel conceptual model of cognitive systems using self-shaping dynamical systems with plastic velocity fields, advancing understanding of biological and artificial cognition.

Findings

Self-shaping DSs can perform cognitive functions without supervision.

The model avoids spurious attractors common in neural networks.

Experimental demonstration with musical examples shows practical potential.

Abstract

Spontaneously evolving living systems can be modelled as continuous-time dynamical systems (DSs), whose evolution rules are determined by their velocity vector fields. We point out that because of their architectural plasticity, biological neural networks belong to a novel type of DSs whose velocity field is plastic, albeit within bounds, and affected by sensory stimuli. We introduce DSs with fully plastic velocity fields self-organising under the influence of stimuli, called self-shaping DSs, and propose that a system of this class represents a conceptual model of a cognitive system. We propose a simple phenomenological model that within a single field-shaping mechanism carries out a set of essential cognitive functions without any supervision and online, just like living cognitive systems do. The performance of this model is illustrated experimentally with musical examples. Unlike in artificial neural networks, this mechanism does not produce spurious attractors and in principle does not limit memory size. Implementation of this principle could pave the way to creating artificial intelligent devices of a new type.