论文标题
生物启发的min-net提高了深网的性能和鲁棒性
Bio-inspired Min-Nets Improve the Performance and Robustness of Deep Networks
论文作者
论文摘要
最小网络的灵感来自其末端的皮质细胞，其单位最少会输出两个学习过滤器。我们将这样的小单元插入最新的深层网络，例如流行的Resnet和Densenet，并表明所得的Min-Nets在CIFAR-10基准上的表现更好。此外，我们表明，针对JPEG压缩工件更强大。我们认为，最小操作是在成对过滤器上实施AN和操作的最简单方法，并且在鉴于自然图像的统计数据的情况下，此类操作和操作引入了适当的偏差。
Min-Nets are inspired by end-stopped cortical cells with units that output the minimum of two learned filters. We insert such Min-units into state-of-the-art deep networks, such as the popular ResNet and DenseNet, and show that the resulting Min-Nets perform better on the Cifar-10 benchmark. Moreover, we show that Min-Nets are more robust against JPEG compression artifacts. We argue that the minimum operation is the simplest way of implementing an AND operation on pairs of filters and that such AND operations introduce a bias that is appropriate given the statistics of natural images.
