Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Personality-guided Public-Private Domain Disentangled Hypergraph-Former Network for Multimodal Depression Detection

Changzeng Fu, Shiwen Zhao, Yunze Zhang, Zhongquan Jian, Shiqi Zhao, Chaoran Liu

TL;DR

Depression detection benefits from modeling individual differences, temporal cross-modal dependencies, and context-specific multi-event information. The authors propose P$^3$HF, combining personality-guided feature regulation, a Hypergraph-Former for high-order temporal fusion, and event-level public-private domain disentanglement with adversarial and HSIC losses. Empirical results on MPDD-Young show substantial improvements in ACC and w-F1 for binary and ternary depression classification, with ablations confirming the necessity of each component. The approach advances personalized, temporally aware multimodal depression analysis and offers a framework potentially applicable to other mental-health detection tasks.

Abstract

Depression represents a global mental health challenge requiring efficient and reliable automated detection methods. Current Transformer- or Graph Neural Networks (GNNs)-based multimodal depression detection methods face significant challenges in modeling individual differences and cross-modal temporal dependencies across diverse behavioral contexts. Therefore, we propose P$^3$HF (Personality-guided Public-Private Domain Disentangled Hypergraph-Former Network) with three key innovations: (1) personality-guided representation learning using LLMs to transform discrete individual features into contextual descriptions for personalized encoding; (2) Hypergraph-Former architecture modeling high-order cross-modal temporal relationships; (3) event-level domain disentanglement with contrastive learning for improved generalization across behavioral contexts. Experiments on MPDD-Young dataset show P$^3$HF achieves around 10\% improvement on accuracy and weighted F1 for binary and ternary depression classification task over existing methods. Extensive ablation studies validate the independent contribution of each architectural component, confirming that personality-guided representation learning and high-order hypergraph reasoning are both essential for generating robust, individual-aware depression-related representations. The code is released at https://github.com/hacilab/P3HF.

Personality-guided Public-Private Domain Disentangled Hypergraph-Former Network for Multimodal Depression Detection

TL;DR

Depression detection benefits from modeling individual differences, temporal cross-modal dependencies, and context-specific multi-event information. The authors propose PHF, combining personality-guided feature regulation, a Hypergraph-Former for high-order temporal fusion, and event-level public-private domain disentanglement with adversarial and HSIC losses. Empirical results on MPDD-Young show substantial improvements in ACC and w-F1 for binary and ternary depression classification, with ablations confirming the necessity of each component. The approach advances personalized, temporally aware multimodal depression analysis and offers a framework potentially applicable to other mental-health detection tasks.

Abstract

Depression represents a global mental health challenge requiring efficient and reliable automated detection methods. Current Transformer- or Graph Neural Networks (GNNs)-based multimodal depression detection methods face significant challenges in modeling individual differences and cross-modal temporal dependencies across diverse behavioral contexts. Therefore, we propose PHF (Personality-guided Public-Private Domain Disentangled Hypergraph-Former Network) with three key innovations: (1) personality-guided representation learning using LLMs to transform discrete individual features into contextual descriptions for personalized encoding; (2) Hypergraph-Former architecture modeling high-order cross-modal temporal relationships; (3) event-level domain disentanglement with contrastive learning for improved generalization across behavioral contexts. Experiments on MPDD-Young dataset show PHF achieves around 10\% improvement on accuracy and weighted F1 for binary and ternary depression classification task over existing methods. Extensive ablation studies validate the independent contribution of each architectural component, confirming that personality-guided representation learning and high-order hypergraph reasoning are both essential for generating robust, individual-aware depression-related representations. The code is released at https://github.com/hacilab/P3HF.

Paper Structure

This paper contains 27 sections, 17 equations, 3 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works
  3. Individual-Aware in Depression Detection
  4. Graph-Based Multimodal Fusion
  5. Disentanglement and Representation Learning
  6. Methodology
  7. Overview.
  8. Preprocessing.
  9. Personality-guided Feature Gating.
  10. Hypergraph-Former Construction.
  11. Public-Private Domain Disentanglement.
  12. Depression Detection.
  13. Experiment
  14. Experiment Protocol
  15. Dataset.
  16. ...and 12 more sections

Figures (3)

  • Figure 1: Model architecture of our P$^3$HF, which is optimized through a combinational objective of three parts. Since the number of events in the dataset we use is three, the legend is drawn in the form of $K=3$.
  • Figure 2: Hyperparameter sensitivity analysis showing optimal configurations for window size and attention heads across binary/ternary tasks. HF:hypergraph-former; HG:hypergraph; CA:cross attention.
  • Figure 3: Domain disentanglement visualization under different loss weights. Optimal configuration ($\beta=\gamma=0.1$) achieves clear public-private feature separation.