Tensorflow Js对层中的量进行偏移_Tensorflow_Neural Network

Tensorflow Js对层中的量进行偏移

tensorflow neural-network

Tensorflow Js对层中的量进行偏移,tensorflow,neural-network,Tensorflow,Neural Network,在一个使用tfjs的简单神经网络设置中，我想知道为什么在一个简单的3隐藏层神经网络中，它在useBias:false和useBias:true之间每层添加了这么多的偏差： _________________________________________________________________ tf-layers.esm.js:2776 Layer (type) Output shape Param # tf-layers.e

在一个使用tfjs的简单神经网络设置中，我想知道为什么在一个简单的3隐藏层神经网络中，它在

useBias:false

和

useBias:true

之间每层添加了这么多的偏差：

_________________________________________________________________ tf-layers.esm.js:2776
Layer (type)                 Output shape              Param #   tf-layers.esm.js:2832
================================================================= tf-layers.esm.js:2776
dense_Dense1 (Dense)         [null,3]                  3         tf-layers.esm.js:2832
_________________________________________________________________ tf-layers.esm.js:2778
dense_Dense2 (Dense)         [null,3]                  9         tf-layers.esm.js:2832
_________________________________________________________________ tf-layers.esm.js:2778
dense_Dense3 (Dense)         [null,3]                  9         tf-layers.esm.js:2832
_________________________________________________________________ tf-layers.esm.js:2778
dense_Dense4 (Dense)         [null,1]                  4         tf-layers.esm.js:2832
================================================================= tf-layers.esm.js:2778
Total params: 25 tf-layers.esm.js:2783
Trainable params: 25 tf-layers.esm.js:2783
Non-trainable params: 0

似乎将

useBias

更改为

true

属性会在同一层中添加许多偏差。最后，这不是很低效吗，因为同一层中的许多偏差最终只会充当一个偏差

________________________________________________________________ tf-layers.esm.js:2776
Layer (type)                 Output shape              Param #   tf-layers.esm.js:2832
================================================================= tf-layers.esm.js:2776
dense_Dense1 (Dense)         [null,3]                  6         tf-layers.esm.js:2832
_________________________________________________________________ tf-layers.esm.js:2778
dense_Dense2 (Dense)         [null,3]                  12        tf-layers.esm.js:2832
_________________________________________________________________ tf-layers.esm.js:2778
dense_Dense3 (Dense)         [null,3]                  12        tf-layers.esm.js:2832
_________________________________________________________________ tf-layers.esm.js:2778
dense_Dense4 (Dense)         [null,1]                  4         tf-layers.esm.js:2832
================================================================= tf-layers.esm.js:2778
Total params: 34 tf-layers.esm.js:2783
Trainable params: 34 tf-layers.esm.js:2783
Non-trainable params: 0