Keras (https://keras.io) is a high-level deep learning framework that works seamlessly with low-level backends like TensorFlow, Theano or CNTK. In Keras a model is like a sequence of layers where each output is fed into the following computational block until the final layer is reached. The generic structure of a model is:
from keras.models import Sequential >>> model = Sequential() >>> model.add(...) >>> model.add(...) ... >>> model.add(...)
The class Sequential
defines a generic empty model, that already implements all the methods needed to add
layers, compile
the model according to the underlying framework, to fit
and evaluate
the model and to predict
the output given an input. All the most common layers are already implemented, including:
- Dense, Dropout and Flattening layers
- Convolutional (1D, 2D and 3D) layers
- Pooling layers
- Zero padding layers
- RNN layers
A model can be compiled using several loss functions (like MSE or cross...