Towards Model-Agnostic Dataset Condensation by Heterogeneous Models

TL;DR

This paper introduces HMDC, a novel dataset condensation method that creates model-agnostic condensed images by leveraging cross-model interactions, addressing the limitations of model-specific synthetic data.

Contribution

The paper proposes a new approach, HMDC, with Gradient Balance Module and Mutual Distillation, to generate universally applicable condensed images across heterogeneous models.

Findings

HMDC produces more versatile condensed datasets.

The Gradient Balance Module improves gradient stability.

Mutual Distillation maintains semantic integrity.

Abstract

Abstract. The advancement of deep learning has coincided with the proliferation of both models and available data. The surge in dataset sizes and the subsequent surge in computational requirements have led to the development of the Dataset Condensation (DC). While prior studies have delved into generating synthetic images through methods like distribution alignment and training trajectory tracking for more efficient model training, a significant challenge arises when employing these condensed images practically. Notably, these condensed images tend to be specific to particular models, constraining their versatility and practicality. In response to this limitation, we introduce a novel method, Heterogeneous Model Dataset Condensation (HMDC), designed to produce universally applicable condensed images through cross-model interactions. To address the issues of gradient magnitude difference and semantic distance in models when utilizing heterogeneous models, we propose the Gradient Balance Module (GBM) and Mutual Distillation (MD) with the SpatialSemantic Decomposition method. By balancing the contribution of each model and maintaining their semantic meaning closely, our approach overcomes the limitations associated with model-specific condensed images and enhances the broader utility. The source code is available in https://github.com/KHU-AGI/HMDC.