Discrete Event, Continuous Time RNNs

TL;DR

This paper introduces a continuous-time extension of the GRU architecture, called CT-GRU, designed to better handle event sequences with varying time scales by incorporating multiple time scales of memory and decay dynamics driven by event timestamps.

Contribution

The paper proposes the CT-GRU, a novel recurrent neural network architecture that integrates continuous-time dynamics and multiple time scales, extending the standard GRU for event-sequence processing.

Findings

CT-GRU performs comparably to standard GRU on multiple datasets.

Incorporating continuous-time dynamics enhances robustness in event-sequence modeling.

Multiple time scales of memory improve handling of diverse temporal patterns.

Abstract

We investigate recurrent neural network architectures for event-sequence processing. Event sequences, characterized by discrete observations stamped with continuous-valued times of occurrence, are challenging due to the potentially wide dynamic range of relevant time scales as well as interactions between time scales. We describe four forms of inductive bias that should benefit architectures for event sequences: temporal locality, position and scale homogeneity, and scale interdependence. We extend the popular gated recurrent unit (GRU) architecture to incorporate these biases via intrinsic temporal dynamics, obtaining a continuous-time GRU. The CT-GRU arises by interpreting the gates of a GRU as selecting a time scale of memory, and the CT-GRU generalizes the GRU by incorporating multiple time scales of memory and performing context-dependent selection of time scales for information storage and retrieval. Event time-stamps drive decay dynamics of the CT-GRU, whereas they serve as generic additional inputs to the GRU. Despite the very different manner in which the two models consider time, their performance on eleven data sets we examined is essentially identical. Our surprising results point both to the robustness of GRU and LSTM architectures for handling continuous time, and to the potency of incorporating continuous dynamics into neural architectures.