Example usage for org.deeplearning4j.nn.multilayer MultiLayerNetwork feedForwardToLayer

Introduction

In this page you can find the example usage for org.deeplearning4j.nn.multilayer MultiLayerNetwork feedForwardToLayer.

Prototype

public List<INDArray> feedForwardToLayer(int layerNum, INDArray input, boolean train) 

Source Link

Document

Compute the activations from the input to the specified layer.
To compute activations for all layers, use feedForward(...) methods
Note: output list includes the original input.

Usage

From source file:org.knime.ext.dl4j.base.nodes.predict.AbstractDLPredictorNodeModel.java

License:Open Source License

/**
 * Activates the specified layer in the specified network with the specified input. The input array should contain
 * one example per row.// w ww .  ja  v a  2 s .co m
 *
 * @param mln the network to use
 * @param layerNum the layer to activate
 * @param input the inputs to use
 * @return the activations of the layer for the input
 */
protected INDArray activate(final MultiLayerNetwork mln, final int layerNum, final INDArray input) {
    /* The MultiLayerNetwork.feedForwardToLayer(int layerNum, INDArray input, boolean train) method is not wrapped
     * into workspaces by DL4J (see MultiLayerNetwork.output(INDArray input, boolean train) method). Therefore, we
     * need to do the same thing here. */
    MemoryWorkspace workspace = Nd4j.getWorkspaceManager()
            .getWorkspaceForCurrentThread(workspaceConfigurationExternal, workspaceExternal);
    try (MemoryWorkspace wsE = workspace.notifyScopeEntered()) {
        final List<INDArray> output = new ArrayList<INDArray>();
        for (int i = 0; i < input.rows(); i++) {
            List<INDArray> activations = mln.feedForwardToLayer(layerNum, input.getRow(i), false);
            output.add(activations.get(activations.size() - 1).detach());
        }
        return Nd4j.hstack(output);
    }
}

---