Neural Network Compression using Transform Coding and Clustering
This addresses the need for efficient neural network compression for mobile and transmission applications, representing an incremental improvement over existing methods.
The paper tackles the problem of neural network file size as a bottleneck for deployment on mobile devices and transmission over limited channels, achieving average compression factors of 7.9-9.3 with only a 1%-2% decrease in accuracy for image classification.
With the deployment of neural networks on mobile devices and the necessity of transmitting neural networks over limited or expensive channels, the file size of the trained model was identified as bottleneck. In this paper, we propose a codec for the compression of neural networks which is based on transform coding for convolutional and dense layers and on clustering for biases and normalizations. By using this codec, we achieve average compression factors between 7.9-9.3 while the accuracy of the compressed networks for image classification decreases only by 1%-2%, respectively.