Non-deterministic Behavior of Ranking-based Metrics when Evaluating Embeddings

TL;DR

This paper analyzes how quantized distances in embedding spaces cause ambiguity in ranking-based performance metrics, affecting system evaluation and security, and proposes a solution to make these metrics deterministic.

Contribution

It provides bounds on the ambiguity caused by quantized distances and introduces a simple method to make ranking-based metrics deterministic and secure against exploits.

Findings

Quantized distances introduce measurable ambiguity in performance metrics.

Security analysis shows potential for exploitation by third parties.

A proposed method makes ranking metrics fully deterministic and secure.

Abstract

Embedding data into vector spaces is a very popular strategy of pattern recognition methods. When distances between embeddings are quantized, performance metrics become ambiguous. In this paper, we present an analysis of the ambiguity quantized distances introduce and provide bounds on the effect. We demonstrate that it can have a measurable effect in empirical data in state-of-the-art systems. We also approach the phenomenon from a computer security perspective and demonstrate how someone being evaluated by a third party can exploit this ambiguity and greatly outperform a random predictor without even access to the input data. We also suggest a simple solution making the performance metrics, which rely on ranking, totally deterministic and impervious to such exploits.