Deep Direct Discriminative Decoders for High-dimensional Time-series Data Analysis

TL;DR

The paper introduces the deep direct discriminative decoder (D4), a neural network-based extension of state-space models that efficiently estimates latent states from high-dimensional time-series data, outperforming traditional methods.

Contribution

It presents a novel formulation of state-space models using deep neural networks to handle high-dimensional and complex observation data.

Findings

D4 outperforms traditional SSMs and RNNs in simulated and real data.

D4 effectively models high-dimensional time-series data with complex distributions.

The approach broadens the applicability of state-space modeling to challenging data types.

Abstract

The state-space models (SSMs) are widely utilized in the analysis of time-series data. SSMs rely on an explicit definition of the state and observation processes. Characterizing these processes is not always easy and becomes a modeling challenge when the dimension of observed data grows or the observed data distribution deviates from the normal distribution. Here, we propose a new formulation of SSM for high-dimensional observation processes. We call this solution the deep direct discriminative decoder (D4). The D4 brings deep neural networks' expressiveness and scalability to the SSM formulation letting us build a novel solution that efficiently estimates the underlying state processes through high-dimensional observation signal. We demonstrate the D4 solutions in simulated and real data such as Lorenz attractors, Langevin dynamics, random walk dynamics, and rat hippocampus spiking neural data and show that the D4 performs better than traditional SSMs and RNNs. The D4 can be applied to a broader class of time-series data where the connection between high-dimensional observation and the underlying latent process is hard to characterize.