Mixed Dimension Embeddings with Application to Memory-Efficient Recommendation Systems
This addresses memory efficiency for large-scale recommendation systems, offering a practical solution to reduce storage needs while preserving performance.
The paper tackles the problem of high memory consumption in embedding layers for recommendation systems by proposing mixed dimension embeddings, where vector dimensions scale with query frequency, achieving a 0.1% accuracy improvement with half the parameters or maintaining accuracy with 16x fewer parameters on the Criteo dataset.
Embedding representations power machine intelligence in many applications, including recommendation systems, but they are space intensive -- potentially occupying hundreds of gigabytes in large-scale settings. To help manage this outsized memory consumption, we explore mixed dimension embeddings, an embedding layer architecture in which a particular embedding vector's dimension scales with its query frequency. Through theoretical analysis and systematic experiments, we demonstrate that using mixed dimensions can drastically reduce the memory usage, while maintaining and even improving the ML performance. Empirically, we show that the proposed mixed dimension layers improve accuracy by 0.1% using half as many parameters or maintain it using 16X fewer parameters for click-through rate prediction task on the Criteo Kaggle dataset.