Keras
Models Model = layers, loss and an optimizer Add layer to model, compile() and fit() Model can be saved and chckpoimted for later use
Layers Layers are used to define the network Examples: Dense Convolutional Pooling Dropout
Loss Function Measures the error by comparing the network predicted output with the expected output The loss must be minimized by udatng the weights thgough backpropagation Common loss functions Mean squared root Cross-entropy
Optimizers Optimizers are strategies used to update the weights durig backpropagation Examples: RMSpro Adam AdaGrad
Backends Keras may use several backends: Exploits either: Theano tensorFlow Exploits either: CPU GPU
Resources https://s3.amazonaws.com/assets.datacamp.com/blog_assets/Keras_Cheat_Sheet_Python.pdf