mapminmax and neural network toolbox

peter

2013 1 18

0 个回答

8 次查看（30 天）

0 个投票

Hi, I trying hard to use the biases and weights of a network build on the toolbox. The function 'mapminmax' is used in the preprocessing and the postprocessing. If the function is turned off the training will performer poorly. If I'm turning off the postprocessing and only use the mapminmax function in the preprocessing, I'm not able to get proper output using the biases and weights in my own program though I'm normalizing the test data.

I have two questions: - why do I not get the right output in my own porgram, when I use the biases and weights, though I'm normalizing the input in the same range, when the toolbox only is normalizing pre-data?

- how or on what elements is the postprocessing normalizing on? (suppose it is on the output from output layer, but can'n see how)

Many thanks in advance!! /Peter