CrescendoNet: A Simple Deep Convolutional Neural Network with Ensemble Behavior
This work addresses the challenge of designing efficient deep learning models for computer vision tasks, offering a novel architecture that reduces memory usage during training, though it appears incremental compared to existing non-residual networks like FractalNet.
The authors tackled the problem of building high-performance deep convolutional neural networks without residual connections by introducing CrescendoNet, which uses simple blocks with independent convolution paths and linearly increasing depths. They found that CrescendoNet with 15 layers and 4.1M parameters outperformed most non-residual networks on CIFAR10, CIFAR100, and SVHN, matching the performance of DenseNet-BC with 250 layers and 15.3M parameters on SVHN.
We introduce a new deep convolutional neural network, CrescendoNet, by stacking simple building blocks without residual connections. Each Crescendo block contains independent convolution paths with increased depths. The numbers of convolution layers and parameters are only increased linearly in Crescendo blocks. In experiments, CrescendoNet with only 15 layers outperforms almost all networks without residual connections on benchmark datasets, CIFAR10, CIFAR100, and SVHN. Given sufficient amount of data as in SVHN dataset, CrescendoNet with 15 layers and 4.1M parameters can match the performance of DenseNet-BC with 250 layers and 15.3M parameters. CrescendoNet provides a new way to construct high performance deep convolutional neural networks without residual connections. Moreover, through investigating the behavior and performance of subnetworks in CrescendoNet, we note that the high performance of CrescendoNet may come from its implicit ensemble behavior, which differs from the FractalNet that is also a deep convolutional neural network without residual connections. Furthermore, the independence between paths in CrescendoNet allows us to introduce a new path-wise training procedure, which can reduce the memory needed for training.