Clustering the Sketch: A Novel Approach to Embedding Table Compression

TL;DR

This paper introduces Clustered Compositional Embeddings (CCE), a novel method for embedding table compression that combines clustering-based quantization with dynamic hashing techniques, enabling efficient training of large recommendation system models.

Contribution

The paper proposes CCE, a new embedding compression method that merges quantization and hashing, with theoretical guarantees of convergence and optimality.

Findings

CCE achieves high compression rates comparable to quantization.

CCE operates dynamically during training, similar to hashing methods.

Theoretical proof guarantees convergence to the optimal codebook.

Abstract

Embedding tables are used by machine learning systems to work with categorical features. In modern Recommendation Systems, these tables can be very large, necessitating the development of new methods for fitting them in memory, even during training. We suggest Clustered Compositional Embeddings (CCE) which combines clustering-based compression like quantization to codebooks with dynamic methods like The Hashing Trick and Compositional Embeddings (Shi et al., 2020). Experimentally CCE achieves the best of both worlds: The high compression rate of codebook-based quantization, but *dynamically* like hashing-based methods, so it can be used during training. Theoretically, we prove that CCE is guaranteed to converge to the optimal codebook and give a tight bound for the number of iterations required.