hiromis/notes

buzzwords

train_loss

valid_loss

error_rate

epoch

overfitting

underfitting

GAN

generative adversarial networks

learning rate

how quickly i can update parameters of my model

Keras is a framework on top of TF and other deep learning libraries.

fast.ai is built on top of PyTorch

one-cycle-learning

fitting a model

how many times does the model go through the training data

Its the same as epoch

validation set

our training model doesn't get to see it

When we are using ImageDataBunch class - it automatically creates a validation set for us.

error metrics are done on validation set. It helps in not doing overfitting