GeForce GTX 1070 (8GB)
ASRock Z170M Pro4S [Intel Z170chipset]
Ubuntu 16.04 LTS desktop amd64
TensorFlow v1.1.0
cuDNN v5.1 for Linux
CUDA v8.0
Python 3.5.2
IPython 6.0.0 -- An enhanced Interactive Python.
gcc (Ubuntu 5.4.0-6ubuntu1~16.04.4) 5.4.0 20160609
GNU bash, version 4.3.48(1)-release (x86_64-pc-linux-gnu)
TensorFlowでの5次元関数の学習に関して。
cost functionやactivation functionsはどういうものが良いか?
Neural Computation: Lecture 7
@ John A. Bullinaria, 2015
http://www.cs.bham.ac.uk/~jxb/INC/l7.pdf
@ L7-9
Regression/Function Approximation Problems
SSE cost function, linear output activations, sigmoid hidden activations
こちらでは現在、上記のcost function, output activation, hidden activationsを使用中。
In each case, application of the gradient descent learning algorithm (by computing the
partial derivatives) leads to appropriate back-propagation weight update equations.
現在AdamOptimizerを使用中。SGDにするとlossの経過は良くなかった。パラメータの選択によるかもしれない。
...
@ L7-16
For regression problems, the SSE cost function itself is often the most useful measure of performance. Sometimes it is helpful to divide that by the number of patterns and take the square root to give the Root-Mean-Squared Error (RMSE). It might also help to normalize the outputs by dividing by the target mean or standard-deviation.
出力層のnormalizeについて書かれていた。
output layerのデータ標準化も試してみたが、まだ誤差は大きい。