In this paper, we introduce a new deep convolutional neural network (ConvNet)
module that promotes competition among a set of multi-scale convolutional
filters. This new module is inspired by the inception module, where we replace
the original collaborative pooling stage (consisting of
本文介绍了 Maxout network In Network (MIN) 这一新型深度学习结构的架构和实现方式,其应用了批量归一化和 dropout 等一系列优化方式,成功解决了在使用 rectifier units 时可能出现的梯度消失问题,同时通过 average pooling 方式进一步提高了信息抽象能力以及遏制了无关信息的干扰,最后在实验数据中取得了卓越的分类表现。