Despite the remarkable success of deep learning in pattern recognition, deep network models face the problem oftraining a large number of parameters. In this paper, we propose and evaluate a novel multi-path wavelet neural networkarchitecture for image classification with far less number of trainable parameters. The model architecture consists of amulti-path layout with several levels of wavelet decompositions performed in parallel followed by fully connectedlayers. These decomposition operations comprise wavelet neurons with learnable parameters, which are updated duringthe training phase using the back-propagation algorithm. We evaluate the performance of the introduced network usingcommon image datasets without data augmentation except for SVHN and compare the results with influential deeplearning models. Our findings support the possibility of reducing the number of parameters significantly in deep neuralnetworks without compromising its accuracy.
展开▼
