DL4J Feedforward Predictor

This Node Is Deprecated — This node is kept for backwards-compatibility, but the usage in new workflows is no longer recommended. The documentation below might contain more information.

This node uses the supplied trained Deep Learning Model to create predictions or activations for the supplied test data. The network output activation will be appended to the training data as a collection column where the collection has the same length as the number of output units of the network, which is usually specified in the Output Layer. The numbers contained in the collection are the raw output activations of the last layer of the network. If specified and the activation of the last layer is 'softmax' the raw activation can be interpreted as class probabilities and be associated with a class label taken from the Deeplearning Model the model was trained on. The supplied data table needs to be in the same format as the table used for learning, meaning it needs to contain columns of the same name and type.


Append softmax prediction?
Whether to append the predicted label to the output table. This is only possible if the output layer of the network configuration uses softmax activation function. The label with the highest corresponding softmax probability will be chosen.
Append Error for each example?
Whether to append the error for each example to the output table. Error is calculated with respect to the set loss function.

Input Ports

Trained Deep Learning Model.
Data table containing testing data.

Output Ports

Data Table containing original testing data with appended network output. The network output is given by a collection column containing output activations for each test example. Additionally, softmax prediction may be appended.


This node has no views




You want to see the source code for this node? Click the following button and we’ll use our super-powers to find it for you.