1. Learnable Heterogeneous Convolution: Learning both topology and strength
- Author
-
Qikun Zhang, Rongzhen Zhao, and Zhenzhi Wu
- Subjects
FOS: Computer and information sciences ,0209 industrial biotechnology ,Artificial neural network ,Computer science ,Cognitive Neuroscience ,Computation ,Computer Vision and Pattern Recognition (cs.CV) ,Computer Science - Computer Vision and Pattern Recognition ,Topology (electrical circuits) ,02 engineering and technology ,Topology ,Convolutional neural network ,Convolution ,020901 industrial engineering & automation ,Kernel (image processing) ,Artificial Intelligence ,0202 electrical engineering, electronic engineering, information engineering ,Code (cryptography) ,Hardware acceleration ,Learning ,020201 artificial intelligence & image processing ,Neural Networks, Computer - Abstract
Existing convolution techniques in artificial neural networks suffer from huge computation complexity, while the biological neural network works in a much more powerful yet efficient way. Inspired by the biological plasticity of dendritic topology and synaptic strength, our method, Learnable Heterogeneous Convolution, realizes joint learning of kernel shape and weights, which unifies existing handcrafted convolution techniques in a data-driven way. A model based on our method can converge with structural sparse weights and then be accelerated by devices of high parallelism. In the experiments, our method either reduces VGG16/19 and ResNet34/50 computation by nearly 5x on CIFAR10 and 2x on ImageNet without harming the performance, where the weights are compressed by 10x and 4x respectively; or improves the accuracy by up to 1.0% on CIFAR10 and 0.5% on ImageNet with slightly higher efficiency. The code will be available on www.github.com/Genera1Z/LearnableHeterogeneousConvolution., Published in Neural Networks journal
- Published
- 2023