One-for-All: Bridge the Gap Between Heterogeneous Architectures in Knowledge Distillation

TL;DR

This paper introduces OFA-KD, a novel knowledge distillation framework that effectively bridges the gap between heterogeneous neural network architectures by projecting features into an aligned space and using adaptive target enhancement.

Contribution

The paper proposes OFA-KD, a simple yet effective method for cross-architecture knowledge distillation that improves performance across CNNs, Transformers, and MLPs.

Findings

Achieves up to 8.0% accuracy gain on CIFAR-100.

Improves 0.7% accuracy on ImageNet-1K.

Demonstrates effectiveness across diverse architectures.

Abstract

Knowledge distillation~(KD) has proven to be a highly effective approach for enhancing model performance through a teacher-student training scheme. However, most existing distillation methods are designed under the assumption that the teacher and student models belong to the same model family, particularly the hint-based approaches. By using centered kernel alignment (CKA) to compare the learned features between heterogeneous teacher and student models, we observe significant feature divergence. This divergence illustrates the ineffectiveness of previous hint-based methods in cross-architecture distillation. To tackle the challenge in distilling heterogeneous models, we propose a simple yet effective one-for-all KD framework called OFA-KD, which significantly improves the distillation performance between heterogeneous architectures. Specifically, we project intermediate features into an aligned latent space such as the logits space, where architecture-specific information is discarded. Additionally, we introduce an adaptive target enhancement scheme to prevent the student from being disturbed by irrelevant information. Extensive experiments with various architectures, including CNN, Transformer, and MLP, demonstrate the superiority of our OFA-KD framework in enabling distillation between heterogeneous architectures. Specifically, when equipped with our OFA-KD, the student models achieve notable performance improvements, with a maximum gain of 8.0% on the CIFAR-100 dataset and 0.7% on the ImageNet-1K dataset. PyTorch code and checkpoints can be found at https://github.com/Hao840/OFAKD.