Confusion Matrix for non-binary classes

10 次查看（过去 30 天）
AlmostSurely
AlmostSurely 2015-5-9
评论： AlmostSurely 2015-5-9
I've trained an FFNN with 10 nodes in the single hidden layer, that takes in 10 inputs, with targets ranging between 0 and 9 (representing the 10 possible classes).
I am trying to plot the confusion matrix using plotconfusion, however, I am getting an error that the numbers are not 0/1. I am new to Matlab and am not sure if such a question is well posed, since network outputs are decimal numbers rather than integers.
I suppose I can round each network output to get integers, but they will still not be 0/1.
EDIT: keyword confusionmat does indeed produce a confusion matrix when I round up network outputs. However, I'd still like to get a plot if possible.
Any help is much appreciated.

请先登录，再回答此问题。

采纳的回答

the cyclist
the cyclist 2015-5-9
I'm a bit confused. (Pun intended.)
The documentation for plotconfusion shows an example that has more than binary classification.
Does that example work for you? If so, maybe you can "interpolate" between that example and your actual problem, to see what went wrong.
  1 个评论
AlmostSurely
AlmostSurely 2015-5-9
Thank you. It seems that I need to use targets of the form (0,0,0,0,0,0,0,0,1) to represent class 9, for example. :)

请先登录，再进行评论。

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Statistics and Machine Learning Toolbox 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