Knowledge Translation: A New Pathway for Model Compression
Authors:
Preprint
Abstract
Deep learning has witnessed significant advancements in recent years at the cost of increasing training, inference, and model storage overhead. While existing model compression methods strive to reduce the number of model parameters while maintaining high accuracy, they inevitably necessitate the re-training of the compressed model or impose architectural constraints. To overcome these limitations, this paper presents a novel framework, termed \textbf{K}nowledge \textbf{T}ranslation (KT), wherein a ``translation'' model is trained to receive the parameters of a larger model and generate compressed parameters. The concept of KT draws inspiration from language translation, which effectively employs neural networks to convert different languages, maintaining identical meaning. Accordingly, we explore the potential of neural networks to convert models of disparate sizes, while preserving their functionality. We propose a comprehensive framework for KT, introduce data augmentation strategies to enhance model performance despite restricted training data, and successfully demonstrate the feasibility of KT on the MNIST dataset. Code is available at \url{https://github.com/zju-SWJ/KT}.Keywords
All Related Versions
This publication has 6 references indexed in Scilit:
- Pruning and quantization for deep neural network acceleration: A surveyNeurocomputing, 2021
- A comprehensive survey on model compression and accelerationArtificial Intelligence Review, 2020
- Generalized Canonical Polyadic Tensor DecompositionSIAM Review, 2020
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998
- A database for handwritten text recognition researchIEEE Transactions on Pattern Analysis and Machine Intelligence, 1994
- On the Early History of the Singular Value DecompositionSIAM Review, 1993