TropNNC: Structured Neural Network Compression Using Tropical Geometry
This addresses the problem of reducing neural network size for deployment, though it appears incremental as it builds on prior geometric approximations.
The paper tackles neural network compression by introducing TropNNC, a framework that uses tropical geometry to represent networks as tropical rational functions and compress them via polynomial reduction, achieving competitive performance on MNIST, CIFAR, and ImageNet while matching strong baselines like ThiNet and CUP.
We present TropNNC, a framework for compressing neural networks with linear and convolutional layers and ReLU activations using tropical geometry. By representing a network's output as a tropical rational function, TropNNC enables structured compression via reduction of the corresponding tropical polynomials. Our method refines the geometric approximation of previous work by adaptively selecting the weights of retained neurons. Key contributions include the first application of tropical geometry to convolutional layers and the tightest known theoretical compression bound. TropNNC requires only access to network weights - no training data - and achieves competitive performance on MNIST, CIFAR, and ImageNet, matching strong baselines such as ThiNet and CUP.