Binarized Neural Networks on the ImageNet Classification Task
This work addresses the challenge of efficient neural network deployment for resource-constrained applications, representing a strong incremental improvement in binarized network performance.
The researchers tackled the problem of training Binarized Neural Networks (BNNs) on the ImageNet classification task, achieving a top-5 accuracy of 84.1% on the validation set, which significantly outperforms previous results like 73.2% and 69.1%.
We trained Binarized Neural Networks (BNNs) on the high resolution ImageNet ILSVRC-2102 dataset classification task and achieved a good performance. With a moderate size network of 13 layers, we obtained top-5 classification accuracy rate of 84.1 % on validation set through network distillation, much better than previous published results of 73.2% on XNOR network and 69.1% on binarized GoogleNET. We expect networks of better performance can be obtained by following our current strategies. We provide a detailed discussion and preliminary analysis on strategies used in the network training.