Parallel Total Variation Distance Estimation with Neural Networks for Merging Over-Clusterings
This provides a domain-independent method for improving clustering results, which is incremental as it builds on existing over-clustering approaches.
The paper tackles the problem of merging over-partitioned clusters in datasets by using total variation distance (TVD) estimated with neural networks, demonstrating better merge decisions than state-of-the-art unsupervised methods on ImageNet subsets and a point cloud dataset.
We consider the initial situation where a dataset has been over-partitioned into $k$ clusters and seek a domain independent way to merge those initial clusters. We identify the total variation distance (TVD) as suitable for this goal. By exploiting the relation of the TVD to the Bayes accuracy we show how neural networks can be used to estimate TVDs between all pairs of clusters in parallel. Crucially, the needed memory space is decreased by reducing the required number of output neurons from $k^2$ to $k$. On realistically obtained over-clusterings of ImageNet subsets it is demonstrated that our TVD estimates lead to better merge decisions than those obtained by relying on state-of-the-art unsupervised representations. Further the generality of the approach is verified by evaluating it on a a point cloud dataset.