Review of compressed embedding layers and their applications for recommender systems

TL;DR

This paper reviews various compressed embedding layer techniques and evaluates their effectiveness in reducing the size of large neural recommender systems.

Contribution

It provides a comprehensive review of compressed embedding methods and reports experimental results demonstrating their practical benefits.

Findings

Compressed embeddings significantly reduce model size.

Experimental results show maintained recommendation accuracy.

Various techniques have different trade-offs in compression and performance.

Abstract

We review the literature on trainable, compressed embedding layers and discuss their applicability for compressing gigantic neural recommender systems. We also report the results we measured with our compressed embedding layers.