How to test a neural network which is trained for multiple input patterns (Own code and not nntool)?

1 次查看（过去 30 天）
Ravi mutturi
Ravi mutturi 2019-7-19
回答： Ravi mutturi 2019-8-3
Hi
I have developed and trained a neural network (3 layers: 1 input, 1 hidden and 1 output) for following situation
(The code was written step by step, as i do not want to jump directly to nntool without understanding the computations)
Data set (40 input patterns):
Input: 40 samples 5 elements
Output: 40 samples 1 element
number of neurons (Input = 5; hidden = 5; output = 1)
Using the delta rule with backpropagation algorithm, i was able to achieve error = 9.39E-06 for 1000 iterations
My final "input to hidden layer" weight matrix size is 200 x 5 (as i have 40 samples x 5 input neurons and 5 hidden neurons)
"hidden to output layer" weight matrix size is 200 x 1 (as i have 40 samples x 5 hidden neurons and 1 output neuron)
Now my question is for a given test sample having 5 elements (input is 1 sample 5 elements), i need to run feed-forward computation to get a single element output.
For running this which weights i need to select in "input to hidden layer" and "hidden to output layer" from the trained set??
I have 200 x 5 and 200 x 1 weight matrices; but i require only 5 x 5 and 5 x 1 weight matrices for testing.
Kindly let me know if i am missing something here?
Thanks in advance
Ravi

请先登录，再回答此问题。

采纳的回答

Ravi mutturi
Ravi mutturi 2019-8-3
There is fundamental mistake in my problem formulation. I am supposed to take same weight matrix for all patterns rather than changing for each pattern. I could fix my code. Thanks for reading the question.

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Pattern Recognition and Classification 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by