Neural Feature Selection for Learning to Rank

TL;DR

This paper introduces a neural feature selection method for learning to rank that significantly reduces input size and model complexity, leading to faster training and inference without sacrificing performance.

Contribution

It proposes an architecture-agnostic neural feature selection approach that cuts input size by up to 60% and reduces model complexity and training/inference time.

Findings

Input size reduced by up to 60%.

Training and inference time decreased by up to 50%.

Performance remains unaffected by feature reduction.

Abstract

LEarning TO Rank (LETOR) is a research area in the field of Information Retrieval (IR) where machine learning models are employed to rank a set of items. In the past few years, neural LETOR approaches have become a competitive alternative to traditional ones like LambdaMART. However, neural architectures performance grew proportionally to their complexity and size. This can be an obstacle for their adoption in large-scale search systems where a model size impacts latency and update time. For this reason, we propose an architecture-agnostic approach based on a neural LETOR model to reduce the size of its input by up to 60% without affecting the system performance. This approach also allows to reduce a LETOR model complexity and, therefore, its training and inference time up to 50%.