Latent Temporal Flows for Multivariate Analysis of Wearables Data

TL;DR

This paper introduces Latent Temporal Flows, a novel multivariate time-series modeling approach for wearable sensor data that improves forecasting accuracy and enables health-related insights using low-dimensional latent representations.

Contribution

The paper presents a new method combining autoencoders and normalizing flows for modeling high-dimensional wearable data, achieving superior forecasting and health indicator identification.

Findings

Achieves at least 10% improvement in multi-step forecasting accuracy.

Effectively identifies VO2 max from low-dimensional latent representations.

Outperforms state-of-the-art methods on real-world wearable datasets.

Abstract

Increased use of sensor signals from wearable devices as rich sources of physiological data has sparked growing interest in developing health monitoring systems to identify changes in an individual's health profile. Indeed, machine learning models for sensor signals have enabled a diverse range of healthcare related applications including early detection of abnormalities, fertility tracking, and adverse drug effect prediction. However, these models can fail to account for the dependent high-dimensional nature of the underlying sensor signals. In this paper, we introduce Latent Temporal Flows, a method for multivariate time-series modeling tailored to this setting. We assume that a set of sequences is generated from a multivariate probabilistic model of an unobserved time-varying low-dimensional latent vector. Latent Temporal Flows simultaneously recovers a transformation of the observed sequences into lower-dimensional latent representations via deep autoencoder mappings, and estimates a temporally-conditioned probabilistic model via normalizing flows. Using data from the Apple Heart and Movement Study (AH&MS), we illustrate promising forecasting performance on these challenging signals. Additionally, by analyzing two and three dimensional representations learned by our model, we show that we can identify participants' $\text{VO}_2\text{max}$, a main indicator and summary of cardio-respiratory fitness, using only lower-level signals. Finally, we show that the proposed method consistently outperforms the state-of-the-art in multi-step forecasting benchmarks (achieving at least a $10\%$ performance improvement) on several real-world datasets, while enjoying increased computational efficiency.