Empirical Evaluation of Rectified Activations in Convolutional Network

RReLU：受限激活函数的经验验证.

Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification

PReLU：分类任务超越人类表现.

Rectifier Nonlinearities Improve Neural Network Acoustic Models

LeakyReLU：使用修正的非线性提高神经网络声学模型.

多任务学习(Multi-Task Learning)

Multi Task Learning.

Training Deeper Convolutional Networks with Deep Supervision

使用深度监督训练更深的卷积网络.

Deeply-Supervised Nets

深度监督网络.