Class KerasLoader


  • public class KerasLoader
    extends java.lang.Object
    • Constructor Summary

      Constructors 
      Constructor Description
      KerasLoader​(java.io.InputStream weightFile, java.io.InputStream configFile)  
    • Method Summary

      All Methods Instance Methods Concrete Methods 
      Modifier and Type Method Description
      MultilayerPerceptron buildModel()
      Keras Weight file format: The weight format is a JSON array every 2 indexes is a layer The 1st of the 2 indexes are the weights.
      • Methods inherited from class java.lang.Object

        equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
    • Constructor Detail

      • KerasLoader

        public KerasLoader​(java.io.InputStream weightFile,
                           java.io.InputStream configFile)
    • Method Detail

      • buildModel

        public MultilayerPerceptron buildModel()
        Keras Weight file format: The weight format is a JSON array every 2 indexes is a layer The 1st of the 2 indexes are the weights. The weights are comprised of 1 JSON array for every input node, and each node contains the weights for all outgoing synapses The 2nd of the 2 indexes are the biases This is just an array of doubles and the length of the bias array is the size of the next layer

        Since as of now, this library will only support MLPs we only care out dense layers If we come across a layer that isn't dense, dropout, or leakyrelu we will throw an error I hope to extend this library one day to be a little more robust

        Returns: