Hey guys im struggling to differentiate between m_train images and m_test images in Logistic_Regression_with_a_Neural_Network_mindset excercise. Can someone explain ?
Generally both train and test data have labels - known correct values. You expose the network to the train set and let it learn weights using it. Then you run forward propagation or predictions on the test set to see how well it does. You don’t want test data to ‘leak’ into training, else the test accuracy isn’t a legitimate measure of how well the algorithm does with data it hasn’t trained on.