Research data related to NeuralNetwork-Experiment2



The dataset supports the thesis "Semantic Impact - A novel approach for domain concept selection in ontology learning". The neural network set trained in the second experiment (by the best NN structure identified as part of the experiment, which contains 3 hidden layers and 1500 nodes on each layer). The format is as follow: TrainedNet_<Concept_Name>_<Source/Target> Source is indicated as Y Target is indicated as X NN is trained by using DeepLearning4J 1.0.0-beta3 and can be loaded with the following code: String strPath = "path to the NN file"; MultiLayerNetwork myNN = ModelSerializer.restoreMultiLayerNetwork(strPath); myNN.getLayerWiseConfigurations().setTrainingWorkspaceMode(WorkspaceMode.ENABLED);
Date made available5 May 2021
PublisherUniversity of Birmingham

Cite this