Time2Vec: Learning a Vector Representation of Time

TL;DR

Time2Vec introduces a model-agnostic vector representation of time that enhances the performance of various models across different applications involving temporal data.

Contribution

The paper proposes Time2Vec, a novel vector representation of time that can be integrated into existing architectures to improve their effectiveness.

Findings

Replacing time with Time2Vec improves model performance

Time2Vec is model-agnostic and easily integrable

Enhances performance across diverse applications

Abstract

Time is an important feature in many applications involving events that occur synchronously and/or asynchronously. To effectively consume time information, recent studies have focused on designing new architectures. In this paper, we take an orthogonal but complementary approach by providing a model-agnostic vector representation for time, called Time2Vec, that can be easily imported into many existing and future architectures and improve their performances. We show on a range of models and problems that replacing the notion of time with its Time2Vec representation improves the performance of the final model.