Derivation of a novel efficient supervised learning algorithm from cortical-subcortical loops

TL;DR

This paper introduces a biologically-inspired supervised learning algorithm derived from cortico-striatal loops, which outperforms traditional machine learning methods like SVM and k-NN in classification tasks, with lower computational costs.

Contribution

The paper presents a novel supervised classification algorithm based on cortical-subcortical loop analysis, demonstrating its efficiency and competitive performance against standard machine learning algorithms.

Findings

Achieves comparable classification accuracy to SVM and k-NN

Operates with significantly reduced time and space complexity

Suggests biological plausibility for associative learning mechanisms

Abstract

Although brain circuits presumably carry out useful perceptual algorithms, few instances of derived biological methods have been found to compete favorably against algorithms that have been engineered for specific applications. We forward a novel analysis of function of cortico-striatal loops, which constitute more than 80% of the human brain, thus likely underlying a broad range of cognitive functions. We describe a family of operations performed by the derived method, including a nonstandard method for supervised classification, which may underlie some forms of cortically-dependent associative learning. The novel supervised classifier is compared against widely-used algorithms for classification, including support vector machines (SVM) and k-nearest neighbor methods, achieving corresponding classification rates --- at a fraction of the time and space costs. This represents an instance of a biologically-derived algorithm comparing favorably against widely used machine learning methods on well-studied tasks.