ITPP: Learning Disentangled Event Dynamics in Marked Temporal Point Processes

TL;DR

ITPP introduces a channel-independent, ODE-based architecture with a type-aware attention mechanism for marked temporal point processes, improving modeling of event type-specific dynamics and outperforming existing methods.

Contribution

The paper proposes ITPP, a novel disentangled, channel-independent MTPP model with a type-aware attention mechanism, addressing entanglement issues in prior models.

Findings

ITPP outperforms state-of-the-art models in predictive accuracy.

ITPP demonstrates improved robustness and generalization.

The architecture effectively captures type-specific event dynamics.

Abstract

Marked Temporal Point Processes (MTPPs) provide a principled framework for modeling asynchronous event sequences by conditioning on the history of past events. However, most existing MTPP models rely on channel-mixing strategies that encode information from different event types into a single, fixed-size latent representation. This entanglement can obscure type-specific dynamics, leading to performance degradation and increased risk of overfitting. In this work, we introduce ITPP, a novel channel-independent architecture for MTPP modeling that decouples event type information using an encoder-decoder framework with an ODE-based backbone. Central to ITPP is a type-aware inverted self-attention mechanism, designed to explicitly model inter-channel correlations among heterogeneous event types. This architecture enhances effectiveness and robustness while reducing overfitting. Comprehensive experiments on multiple real-world and synthetic datasets demonstrate that ITPP consistently outperforms state-of-the-art MTPP models in both predictive accuracy and generalization.