Test a model
The test-model function helps measuring a model's performance at the recognition task, given a dataset. To do so, for each sample given in input, it compares the actual class and predicted class. Counting the errors and the successes, an accuracy is output.
Plug the trained xmm-model object's self output into the model input of the test-model function.
The Dataset input expects the exact same format as the dataset input of the xmm object : a list of labelled samples.
The dataset used for training can be used for testing by plugging in the dataset output of the xmm-model object.
The accuracy measure (between 0 and 1) is output by the function when evaluated. 1 means the model has successfully classified every sample, 0 means it didn't predict the right class of any sample.
To better understand the model's behaviour, the xmm-model object stores the errors and successes encountered throughout testing. The user can access these informations with two distinct functions, after the test-model function has been evaluated.
For both of them, plug the tested xmm-model object's self output into the self input.
Use get-errors and get-confusion-matrix to display your tests' detailled results
Get-errors prints in the console the number of errors for each label to predict.
The get-confusion-matrix function outputs a 2d-array object containing the confusion matrix of the test.
Plug its output to the self input of a 2d-array object to display the matrix.
The rows correspond to the actual classes, and the column the predicted classes. Thus, 0.17 at "A"x"B" indicates that 17% of the samples of the "A" class were classified as "B". The numbers on the right column of the matrix indicate the total number of samples for each class.