A Greedy Algorithm to Cluster Specialists
This work addresses a specific bottleneck in deep neural network experiments for researchers, but it is incremental as it modifies existing clustering procedures.
The paper tackled the problem of clustering algorithms for class assignment in the generalist-specialist paradigm for classification, finding that their proposed greedy pairs algorithm consistently outperforms other alternatives on CIFAR-10 and CIFAR-100 datasets.
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.