# Binary Latent Representations for Efficient Ranking: Empirical Assessment

## TL;DR

This paper evaluates the trade-offs of using binary latent representations in large-scale recommender systems, finding that traditional low-dimensional models outperform binary models in accuracy-speed balance.

## Contribution

It provides an empirical assessment of binary latent representations, demonstrating their limitations compared to low-dimensional real-valued models.

## Key findings

- Binary representations are faster but less accurate.
- Low-dimensional models outperform binary models in accuracy.
- Binary models have significant accuracy loss at speed gains.

## Abstract

Large-scale recommender systems often face severe latency and storage constraints at prediction time. These are particularly acute when the number of items that could be recommended is large, and calculating predictions for the full set is computationally intensive. In an attempt to relax these constraints, we train recommendation models that use binary rather than real-valued user and item representations, and show that while they are substantially faster to evaluate, the gains in speed come at a large cost in accuracy. In our Movielens 1M experiments, we show that reducing the latent dimensionality of traditional models offers a more attractive accuracy/speed trade-off than using binary representations.