Deep clustering with associative memories
Loading...
Authors
Saha, Bishwajit
Issue Date
2025-05
Type
Electronic thesis
Thesis
Thesis
Language
en_US
Keywords
Computer science
Alternative Title
Abstract
Clustering is a widely used unsupervised learning technique involving an intensive discrete optimization problem. Recently, there has been growing interest in making clustering end-to-end differentiable. Associative Memory models or AMs are differentiable neural networks defining a recursive dynamical system, which have been integrated with various deep learningarchitectures. However, until recently, there has been no work utilizing Associative Memory models for clustering. In this thesis, we discuss our contributions toward developing an Associative Memory-based clustering scheme. We introduce three innovations that leverage AM models, making them suitable for clustering while preserving their dynamic nature and discrete assignment properties. We first uncover a novel connection between the AM dynamics and the inherent discrete assignment necessary in clustering and then propose a novel unconstrained continuous relaxation of the discrete clustering problem, enabling End-to-end Differentiable Clustering with AM, dubbed ClAM. Leveraging the pattern completion ability of AMs, we further develop a novel self-supervised clustering loss. Our evaluations on varied datasets demonstrate that ClAM benefits from the self-supervision, and significantly improves upon both the traditional Lloyd’s k-means algorithm, and more recent continuous clustering relaxations (by upto 60% in terms of the Silhouette Coefficient). Our second contribution aims to extend ClAM by integrating autoencoder and associative memories together to perform clustering in the latent space instead of the ambient space. Deep clustering, which is joint representation learning and latent space clustering, is a well-studied problem especially in computer vision and text processing under the deep learning framework. While representation learning is generally differentiable, clustering is an inherently discrete optimization task, requiring various approximations and regularizations to fit in a standard differentiable pipeline. This leads to a somewhat disjointed representation learning and clustering. We show how Associative Memories enable a novel take on deep clustering, called DClAM, which simplifies the whole pipeline and ties together the representation learning and clustering more intricately. Our experiments showcase the advantage of DClAM, producing improved clustering quality regardless of the autoencoder architecture choice (convolutional, residual or fully-connected) or data modality (images or text).
Description
May2025
School of Science
School of Science
Full Citation
Publisher
Rensselaer Polytechnic Institute, Troy, NY