Course1 - Week 2 - PA 2 - Logistic Regression with a Neural Network mindset

[nucleosome]

In part 2 where we try to reshape the training and test examples, i.e. to flatten matrix "train_set_x_orig" and "test_set_x_orig".

In this note, the author uses "train_set_x_flatten = train_set_x_orig.reshape(train_set_x_orig.shape[1]*train_set_x_orig.shape[2]*train_set_x_orig.shape[3],train_set_x_orig.shape[0])", which fails in making every colum as a training example and a wrong result of train accuracy and test accuracy.

I think "train_set_x_flatten = train_set_x_orig.reshape(train_set_x_orig.shape[0], -1).T" would be better.

please feel free to discuss this:)

[BJBijvoet]

You are correct, the training and testing examples should be reshaped as:

train_set_x_flatten = train_set_x_orig.reshape(train_set_x_orig.shape[0], -1).T
test_set_x_flatten = test_set_x_orig.reshape(test_set_x_orig.shape[0], -1).T

[woo1127]

A more shorter answer is

train_set_x_flatten = train_set_x_orig.reshape(-1, train_set_x_orig.shape[0])
test_set_x_flatten = test_set_x_orig.reshape(-1, test_set_x_orig.shape[0])