A crossover code for high-dimensional composition

TL;DR

This paper introduces a novel encoding method called crossover codes for high-dimensional vectors, inspired by genetic crossover, enabling efficient, overlapping, and robust compositional representations suitable for fast decoding and flexible information processing.

Contribution

The paper proposes a new crossover encoding scheme for high-dimensional vectors that preserves relational information and allows efficient decoding, differing from existing HD computing methods.

Findings

Crossover codes maintain high overlap with base elements and sub-structures.

The method enables fast greedy decoding of compositional information.

Crossover encoding is mathematically tractable and robust.

Abstract

We present a novel way to encode compositional information in high-dimensional (HD) vectors. Inspired by chromosomal crossover, random HD vectors are recursively interwoven, with a fraction of one vector's components masked out and replaced by those from another using a context-dependent mask. Unlike many HD computing schemes, "crossover" codes highly overlap with their base elements' and sub-structures' codes without sacrificing relational information, allowing fast element readout and decoding by greedy reconstruction. Crossover is mathematically tractable and has several properties desirable for robust, flexible representation.