IEEE Trans Pattern Anal Mach Intell. 2018 May;40(5):1245-1258. doi: 10.1109/TPAMI.2017.2701831. Epub 2017 May 5.
Recent years have witnessed the success of deep neural networks in dealing with a plenty of practical problems. Dropout has played an essential role in many successful deep neural networks, by inducing regularization in the model training. In this paper, we present a new regularized training approach: Shakeout. Instead of randomly discarding units as Dropout does at the training stage, Shakeout randomly chooses to enhance or reverse each unit's contribution to the next layer. This minor modification of Dropout has the statistical trait: the regularizer induced by Shakeout adaptively combines , and regularization terms. Our classification experiments with representative deep architectures on image datasets MNIST, CIFAR-10 and ImageNet show that Shakeout deals with over-fitting effectively and outperforms Dropout. We empirically demonstrate that Shakeout leads to sparser weights under both unsupervised and supervised settings. Shakeout also leads to the grouping effect of the input units in a layer. Considering the weights in reflecting the importance of connections, Shakeout is superior to Dropout, which is valuable for the deep model compression. Moreover, we demonstrate that Shakeout can effectively reduce the instability of the training process of the deep architecture.
近年来，深度神经网络在处理大量实际问题方面取得了成功。在模型训练中引入正则化，Dropout 在许多成功的深度神经网络中发挥了重要作用。本文提出了一种新的正则化训练方法：Shakeout。Shakeout 不是像 Dropout 那样在训练阶段随机丢弃单元，而是随机选择增强或反转每个单元对下一层的贡献。这种对 Dropout 的微小修改具有统计特性：Shakeout 诱导的正则化项自适应地组合了 和正则化项。我们在图像数据集 MNIST、CIFAR-10 和 ImageNet 上使用代表性的深度架构进行分类实验表明，Shakeout 有效地处理了过拟合问题，并且优于 Dropout。我们在无监督和监督设置下都经验性地证明了 Shakeout 会导致权重更加稀疏。Shakeout 还会导致层中输入单元的分组效果。考虑到权重可以反映连接的重要性，Shakeout 优于 Dropout，这对于深度模型压缩很有价值。此外，我们还证明了 Shakeout 可以有效地降低深度架构训练过程的不稳定性。
