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.

Interactive visual query of density maps on latent space via flow‐based models

A Discrete Probabilistic Approach to Dense Flow Visualization

Adaptively Exploring Population Mobility Patterns in Flow Visualization

Scalable Lagrangian-Based Attribute Space Projection for Multivariate Unsteady Flow Data

Integration-free Learning of Flow Maps

Advection-Based Sparse Data Management For Visualizing Unsteady Flow

A Quick and Feature based Visualization Algorithm for Large-scale Flow Data.

Depth-Based Feature Enhancement for Volume Visualization

Image-Space Texture-Based Output-Coherent Surface Flow Visualization.

Let It Flow: A Static Method for Exploring Dynamic Graphs

Computing high-dimensional optimal transport by flow neural networks

Latent Space Cartography: Visual Analysis of Vector Space Embeddings

Latent Variable Modelling with Hyperbolic Normalizing Flows

Spatiotemporal Flow L-function: a new method for identifying spatiotemporal clusters in geographical flow data

Visualizing Large-Scale and High-Dimensional Data

Latent Temporal Flows for Multivariate Analysis of Wearables Data

Density-aware Stratified Sampling for Visualizing Large Volume Geo-Spatial Data

Interactive Volume Visualization via Multi-Resolution Hash Encoding based Neural Representation

DELFlow: Dense Efficient Learning of Scene Flow for Large-Scale Point Clouds

A Tale of Two Latent Flows: Learning Latent Space Normalizing Flow with Short-run Langevin Flow for Approximate Inference

An Adversarial Non-Volume Preserving Flow Model with Boltzmann Priors