Lets keep it simple, Using simple architectures to outperform deeper and more complex architectures

22 Aug 2016Seyyed Hossein Hasanpour • Mohammad Rouhani • Mohsen Fayyaz • Mohammad Sabokrou

We propose a simple architecture, called SimpleNet, based on a set of designing principles, with which we empirically show, a well-crafted yet simple and reasonably deep architecture can perform on par with deeper and more complex architectures. SimpleNet provides a good tradeoff between the computation/memory efficiency and the accuracy. Our simple 13-layer architecture outperforms most of the deeper and complex architectures to date such as VGGNet, ResNet, and GoogleNet on several well-known benchmarks while having 2 to 25 times fewer number of parameters and operations.

Full paper

Evaluation


Task Dataset Model Metric name Metric value Global rank Compare
Image Classification CIFAR-10 SimpleNetv1 Percentage correct 95.51 # 11
Image Classification CIFAR-100 SimpleNetv1 Percentage correct 78.37 # 6
Image Classification MNIST SimpleNetv1 Percentage error 0.2 # 2