FWU 0.0.2

FWU is a series of PyTorch-based neural network units. It has higher information utilization, prevents neuronal necrosis, and eliminates the need to add additionall activation functions.

FWU

FWU is a series of PyTorch-based neural network units.It has higher information utilization, prevents neuronal necrosis, and eliminates the need to add additionall activation functions.

思路说明

ReLU滤除了小于零的输入,虽然高效实现了非线性,但对小于零的输入无梯度回传,导致了神经元坏死.目前有多种模仿ReLU的平滑版激活函数,通过在输入刚小于零时提供显著的反向梯度、在梯度更小时也提供微小的梯度,部分缓解了ReLU中神经元坏死的问题. FWU系列直接输入的符号将输入分为>0和<0两部分,分别送入两个单元(Linear或Conv或其他),再对两个单元的输出做二元运算(加法或乘法或其他).即提供了非线性,又最大程度上利用信息完全杜绝了神经元坏死.

目前提供了加法(FWLU)和乘法(FWMU)的Linear、Conv1d、Conv2d、和对应转置卷积的实现.无后缀代表Linear,1D、2D后缀代表对应维度的卷积,加T后缀代表转置卷积.

Install

pip install FWU

Use

from FWU import FWLU2DT
...

import后直接替代对应的Linear、Conv1d、Conv2d等,并取消与其组合的激活函数(输出层的SoftMax等承担特殊功能的激活函数除外).

HomePage

https://github.com/PsycheHalo/FWU/