Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- Paypal Tensorflow Talk
- Evaluating output accuracy with cost function
- E.g. label = 1, output probability = 0.68, so cost = 0.32. Kind of measuring loss. When training a neural network, we are trying to minimize this cost by picking weights and biases.
- Most used cost function: Cross entropy. Gradient descent to achieve this minimization goal.
Add Comment
Please, Sign In to add comment