Shared Weights Embedding Layer (Stack / Unstack)

thomas · July 2, 2021, 7:10pm

Hi,
i want to get a model nearly like your test case:

eclipse/deeplearning4j/blob/6335b5e933f13d2bdb5c1f7e5896c45eb96839d7/deeplearning4j-core/src/test/java/org/deeplearning4j/nn/graph/graphnodes/TestGraphNodes.java#L295-L304


      
          ComputationGraphConfiguration conf = new NeuralNetConfiguration.Builder().graphBuilder().addInputs("in1", "in2")
                          .addVertex("stack", new org.deeplearning4j.nn.conf.graph.StackVertex(), "in1", "in2")
                          .addLayer("1", new EmbeddingLayer.Builder().nIn(5).nOut(5).build(), "stack")
                          .addVertex("unstack1", new org.deeplearning4j.nn.conf.graph.UnstackVertex(0, 2), "1")
                          .addVertex("unstack2", new org.deeplearning4j.nn.conf.graph.UnstackVertex(0, 2), "1")
                          .addLayer("out1", new OutputLayer.Builder().activation(Activation.TANH)
                                          .lossFunction(LossFunctions.LossFunction.L2).nIn(5).nOut(5).build(), "unstack1")
                          .addLayer("out2", new OutputLayer.Builder().activation(Activation.TANH)
                                          .lossFunction(LossFunctions.LossFunction.L2).nIn(5).nOut(5).build(), "unstack2")
                          .setOutputs("out1", "out2").build();

my question: ist every unstack vertex like a pop operation to the given stack. i would like to merge every result from the embedded layer to one big output, which goes to one single target layer.
like

addVertex("unstack1", new UnstackVertex(0, 2),"embedded")
addVertex("unstack2", new UnstackVertex(0, 2),"embedded")
addVertex("merge", new MergeVertex(), "unstack1","unstack2")

Best regards

Thomas

Topic		Replies	Views
Querying MultiLayerNetwork for configuration properties DL4J	4	324	January 7, 2021
How can I program this neural network architecture in Deeplearning4j? DL4J	4	1052	September 18, 2020
Questions on DL4J application DL4J	3	316	March 24, 2021
ComputationGraph or non-sequential example? DL4J	6	294	January 5, 2022
how to structure multidataset format to train ComputationGraph in deeplearning4j? DL4J	5	735	July 8, 2020

Shared Weights Embedding Layer (Stack / Unstack)

Related Topics