Universal hidden monotonic trend estimation with contrastive learning

TL;DR

This paper introduces a universal contrastive learning method, CTE, capable of extracting hidden monotonic trends from diverse time series data without standard assumptions.

Contribution

The paper presents CTE, a novel contrastive learning approach that detects monotonic trends across various data types, expanding beyond traditional methods.

Findings

CTE successfully identifies hidden monotonic trends in diverse data types.

CTE outperforms traditional trend detection methods in flexibility and applicability.

Experiments demonstrate CTE's effectiveness on vector, image, graph, and time series data.

Abstract

In this paper, we describe a universal method for extracting the underlying monotonic trend factor from time series data. We propose an approach related to the Mann-Kendall test, a standard monotonic trend detection method and call it contrastive trend estimation (CTE). We show that the CTE method identifies any hidden trend underlying temporal data while avoiding the standard assumptions used for monotonic trend identification. In particular, CTE can take any type of temporal data (vector, images, graphs, time series, etc.) as input. We finally illustrate the interest of our CTE method through several experiments on different types of data and problems.