A multi-layer network based on Sparse Ternary Codes for universal vector compression

TL;DR

This paper introduces a multi-layer Sparse Ternary Codes system for efficient vector compression and similarity search, demonstrating improved rate-distortion performance over existing binary hashing methods on synthetic and real data.

Contribution

The paper extends Sparse Ternary Codes to multiple layers, creating ML-STC, which refines vector reconstruction and enhances compression quality while maintaining fast search capabilities.

Findings

ML-STC outperforms single-layer codes in rate-distortion trade-offs.

ML-STC surpasses several binary hashing methods on large-scale datasets.

The multi-layer approach effectively refines residuals for better reconstruction.

Abstract

We present the multi-layer extension of the Sparse Ternary Codes (STC) for fast similarity search where we focus on the reconstruction of the database vectors from the ternary codes. To consider the trade-offs between the compactness of the STC and the quality of the reconstructed vectors, we study the rate-distortion behavior of these codes under different setups. We show that a single-layer code cannot achieve satisfactory results at high rates. Therefore, we extend the concept of STC to multiple layers and design the ML-STC, a codebook-free system that successively refines the reconstruction of the residuals of previous layers. While the ML-STC keeps the sparse ternary structure of the single-layer STC and hence is suitable for fast similarity search in large-scale databases, we show its superior rate-distortion performance on both model-based synthetic data and public large-scale databases, as compared to several binary hashing methods.