Problem with "Confusion" in neural network toolbox

1 次查看（过去 30 天）
D C
D C 2013-10-14
评论： Greg Heath 2013-10-18
I am using matlab r2009b and implementing classification problem (mentioned here http://www.mathworks.com/matlabcentral/answers/89395-early-stopping-method-in-neural-networks-digits-recognition). During training, after validation stop occurs, I try to see confusion plots and I get error "??? Error using ==> confusion at 60 Targets are not all 1/0 values." As I said in a link above tagets are values from 1-10, how can i fix it? Morover, I tried using nprtool and I cannot fill space for targets
Any idea how to fix both problems? Thanks in advance

请先登录，再回答此问题。

采纳的回答

Greg Heath
Greg Heath 2013-10-16
编辑：Greg Heath 2013-10-16
If you have a c-class classifier, the target matrix should contain columns of the c-dimensional unit matrix eye(c).
If the output transfer function is softmax, logsig or the unfortunate tansig default, the output matrix components should be in the interval (0,1).
If the output transfer function is purelin, there may be outputs outside the [ 0,1 ]interval. That should not affect classification if compet and/or vec2ind is used
help compet
help vec2ind
Thank you for formally accepting my answer
Greg

更多回答（1 个）

D C
D C 2013-10-17
I used ind2vec and now my targets are 10x5000 matrix instead of 1x5000, but still doesn't work "N-dimensional indexing allowed for Full matrices only."

类别

在 Help Center 和 File Exchange 中查找有关 Define Shallow Neural Network Architectures 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