A Greedy Algorithm to Cluster Specialists

TL;DR

This paper evaluates clustering algorithms for specialist-generalist neural network classification, proposing a new greedy pairs method that improves performance on CIFAR datasets.

Contribution

It introduces a novel greedy pairs clustering algorithm optimized for specialist-generalist systems, with modifications to existing methods and comprehensive experimental validation.

Findings

Greedy pairs clustering outperforms other algorithms.

Confusion matrix choice has minimal impact on results.

Effective for varying class numbers on CIFAR datasets.

Abstract

Several recent deep neural networks experiments leverage the generalist-specialist paradigm for classification. However, no formal study compared the performance of different clustering algorithms for class assignment. In this paper we perform such a study, suggest slight modifications to the clustering procedures, and propose a novel algorithm designed to optimize the performance of of the specialist-generalist classification system. Our experiments on the CIFAR-10 and CIFAR-100 datasets allow us to investigate situations for varying number of classes on similar data. We find that our \emph{greedy pairs} clustering algorithm consistently outperforms other alternatives, while the choice of the confusion matrix has little impact on the final performance.