Confusion matrix and test accuracy for PyTorch CNN tutorial

Question

I am interested in reporting only train and test accuracy as well as confusion matrix (say using sklearn confusionmatrix). How can I do that? The current tutorial only reports train/val accuracy and I am having hard time figuring how to incorporate the sklearn confusionmatrix code there. Link to original tutorial here: https://github.com/bentrevett/pytorch-sentiment-analysis/blob/master/4%20-%20Convolutional%20Sentiment%20Analysis.ipynb

Show your efforts and code then ask about where is the bug or error? — Muhammad Faizan Fareed

islijepcevic islijepcevic · Accepted Answer · 2020-04-07T09:05:34

Very similar to the binary_accuracy function defined in the tutorial, you can implement any metric that you want. All you need is a set of model predictions (preds in this case) and true targets (y).

For example, for confusion matrix, you could do the following:

from sklearn.metrics import confusion_matrix

def compute_confusion_matrix(preds, y):
    #round predictions to the closest integer
    rounded_preds = torch.round(torch.sigmoid(preds))
    return confusion_matrix(y, rounded_preds)

Confusion matrix and test accuracy for PyTorch CNN tutorial

1 Answers