Neural network-based clustering using pairwise constraints

TL;DR

This paper introduces a neural network framework for clustering that uses pairwise constraints and contrastive learning, eliminating the need for predefined cluster centers or distance metrics, and outperforms traditional methods.

Contribution

The proposed end-to-end neural clustering approach leverages weak pairwise labels and contrastive criteria, providing a data-driven clustering without explicit cluster centers or distance metrics.

Findings

Outperforms conventional two-stage clustering methods.

Robust to the number of clusters, accurately identifying dominant clusters.

Effective even with large numbers of clusters k.

Abstract

This paper presents a neural network-based end-to-end clustering framework. We design a novel strategy to utilize the contrastive criteria for pushing data-forming clusters directly from raw data, in addition to learning a feature embedding suitable for such clustering. The network is trained with weak labels, specifically partial pairwise relationships between data instances. The cluster assignments and their probabilities are then obtained at the output layer by feed-forwarding the data. The framework has the interesting characteristic that no cluster centers need to be explicitly specified, thus the resulting cluster distribution is purely data-driven and no distance metrics need to be predefined. The experiments show that the proposed approach beats the conventional two-stage method (feature embedding with k-means) by a significant margin. It also compares favorably to the performance of the standard cross entropy loss for classification. Robustness analysis also shows that the method is largely insensitive to the number of clusters. Specifically, we show that the number of dominant clusters is close to the true number of clusters even when a large k is used for clustering.