论文标题
COSHNET：使用剪切物的混合复合物有价值的神经网络
CoShNet: A Hybrid Complex Valued Neural Network using Shearlets
论文作者
论文摘要
在混合神经网络中，昂贵的卷积层被不可训练的固定变换所取代，参数大幅减少。在以前的作品中，通过用小波代替卷积来获得良好的结果。但是，基于小波的混合网络遗传了小波沿曲线及其轴偏置的消失力矩。我们建议将剪力夹在边缘，山脊和斑点等重要图像特征（如重要图像）方面提供强大的支持。所得网络称为复杂的剪切网络（COSHNET）。它在针对Resnet-50和Resnet-18的时装摄影师上进行了测试，分别获得了92.2％，分别为90.7％和91.8％。所提出的网络具有49.9k参数，而RESNET-18的参数为11.18m，使用少52倍。最后，我们在Resnet要求的200个时期与200个时期进行了培训，并且不需要任何高参数调整或正则化。 代码：https：//github.com/ujjawal-k-panchal/coshnet
In a hybrid neural network, the expensive convolutional layers are replaced by a non-trainable fixed transform with a great reduction in parameters. In previous works, good results were obtained by replacing the convolutions with wavelets. However, wavelet based hybrid network inherited wavelet's lack of vanishing moments along curves and its axis-bias. We propose to use Shearlets with its robust support for important image features like edges, ridges and blobs. The resulting network is called Complex Shearlets Network (CoShNet). It was tested on Fashion-MNIST against ResNet-50 and Resnet-18, obtaining 92.2% versus 90.7% and 91.8% respectively. The proposed network has 49.9k parameters versus ResNet-18 with 11.18m and use 52 times fewer FLOPs. Finally, we trained in under 20 epochs versus 200 epochs required by ResNet and do not need any hyperparameter tuning nor regularization. Code: https://github.com/Ujjawal-K-Panchal/coshnet
