Keras and DL4J models

DL4J and Keras both have two model types. In DL4J, a simple model constructed by linearly stacking layers is called a MultiLayerNetwork, whereas in Keras this kind of model is called Sequential. More complex neural networks, for instance networks with many inputs, are built as ComputationGraphs in DL4J, while Keras uses the so called Model API.

Keras model import maps Sequential models to MultiLayerNetworks and Models to ComputationGraphs.


Chat with us on Gitter