Examples of ca.nengo.model.Network

ca.nengo.model.Network

A neural circuit, consisting of Nodes such as Ensembles and ExternalInputs. A Network is the usual object of a simulation. If you are new to this code, what you probably want to do is create some Neurons, group them into Ensembles, connect the Ensembles in a Network, and run the Network in a Simulator.

Note: Multiple steps are needed to add a Projection between Ensembles. First, an Origin must be created on the presynaptic Ensemble, and a Termination with the same dimensionality must be created on the post-synaptic Ensemble. Then the Origin and Termination can be connected with the method addProjection(Origin, Termination). We don't do this in one step (ie automatically create the necessary Origin and Termination as needed) because there are various ways of doing so, and in fact some types of Origins and Terminations can only be created in the course of constructing the Ensemble. Creation of an Origin or Termination can also be a complex process. Rather than try to abstract these varied procedures into something that can be driven from the Network level, we just assume here that the necessary Origins and Terminations exist, and provide a method for connecting them.
@author Bryan Tripp

   * 
   * @throws StructuralException 
   * @throws SimulationException 
   */
  public void functionalTestSort() throws StructuralException, SimulationException {
    Network network = new NetworkImpl();
    
    FunctionInput input = new FunctionInput("input", new Function[]{new SineFunction(5)}, Units.UNK);
    network.addNode(input);
    
    NEFEnsembleFactory ef = new NEFEnsembleFactoryImpl();
    NEFEnsemble ensemble = ef.make("ensemble", 100, 1);
    ensemble.addDecodedTermination("input", MU.I(1), .005f, false);
    ensemble.collectSpikes(true);
    network.addNode(ensemble);
    
    network.addProjection(input.getOrigin(FunctionInput.ORIGIN_NAME), ensemble.getTermination("input"));
    network.run(0, 2);
    
    SpikePattern unsorted = ensemble.getSpikePattern();
    SpikePattern sorted = DataUtils.sort(unsorted, ensemble);
    
    Plotter.plot(unsorted);

View Full Code Here

        try {
            NEFEnsemble ensemble = ef.make("test", 20, 1);
            ensemble.addDecodedTermination("input", MU.I(1), .01f, false);
            Plotter.plot(ensemble);


            Network network = new NetworkImpl();
            network.addNode(ensemble);
            FunctionInput input = new FunctionInput("input", new Function[]{new SineFunction(3)}, Units.UNK);
            network.addNode(input);
            network.addProjection(input.getOrigin(FunctionInput.ORIGIN_NAME), ensemble.getTermination("input"));


            network.setMode(SimulationMode.RATE);
            Probe rates = network.getSimulator().addProbe("test", "rate", true);
            network.run(0, 2);
            //          Plotter.plot(rates.getData(), .05f, "rates");
            Plotter.plot(rates.getData(), "rates");
        } catch (StructuralException e) {
            e.printStackTrace();
        } catch (SimulationException e) {

View Full Code Here


public class IntegratorExample {


  public static Network createNetwork() throws StructuralException {
    
    Network network = new NetworkImpl();
    
    Function f = new ConstantFunction(1, 1f);
//    Function f = new SineFunction();
    FunctionInput input = new FunctionInput("input", new Function[]{f}, Units.UNK);
    network.addNode(input);
    
    NEFEnsembleFactory ef = new NEFEnsembleFactoryImpl();
    
    NEFEnsemble integrator = ef.make("integrator", 500, 1, "integrator1", true);  
    network.addNode(integrator);
    integrator.collectSpikes(true);


    Plotter.plot(integrator);
    Plotter.plot(integrator, NEFEnsemble.X);
    
    float tau = .05f; 
    
    Termination interm = integrator.addDecodedTermination("input", new float[][]{new float[]{tau}}, tau, false);
//    Termination interm = integrator.addDecodedTermination("input", new float[][]{new float[]{1f}}, tau);
    network.addProjection(input.getOrigin(FunctionInput.ORIGIN_NAME), interm);
    
    Termination fbterm = integrator.addDecodedTermination("feedback", new float[][]{new float[]{1f}}, tau, false);
    network.addProjection(integrator.getOrigin(NEFEnsemble.X), fbterm);
    
    //System.out.println("Network creation: " + (System.currentTimeMillis() - start));
    return network;
  }

View Full Code Here

    
  
  public static void main(String[] args) {
        
    try {
      Network network = createNetwork();
      Simulator simulator = network.getSimulator();
      
      Probe inputRecorder = simulator.addProbe("input", "input", true);
      Probe integratorRecorder = simulator.addProbe("integrator", NEFEnsemble.X, true);
      Probe neuronRecorder = simulator.addProbe("integrator", 0, "V", true);


      long startTime = System.currentTimeMillis();
      simulator.run(0f, 1f, .0002f);
      System.out.println("Run time: " + ((System.currentTimeMillis() - startTime)/1000f) );
    
      TimeSeries integratorData = integratorRecorder.getData();
      integratorData.getLabels()[0] = "decoded output";
      
      Plotter.plot(inputRecorder.getData(), "Input");
      Plotter.plot(integratorData, .005f, "Integrator");
      Plotter.plot(neuronRecorder.getData(), "Neuron #0");
      
      Plotter.plot(((Ensemble) network.getNode("integrator")).getSpikePattern());
      
      MatlabExporter me = new MatlabExporter();
      me.add("input", inputRecorder.getData());
      me.add("integrator", integratorRecorder.getData(), .01f);
      me.add("neuron", neuronRecorder.getData());

View Full Code Here

public class ScopeExample {


  public static void main(String[] args) throws Exception {
    
    try {
      Network network = new NetworkImpl();
      
      //this "function input" is Probeable ...     
      Function[] funcs = new Function[2];
      for(int i=0; i<funcs.length; ++i) {
        funcs[i] = new PostfixFunction("sin(x0)^"+(i+1), 1);
      }
      String name = "functions of time";
      FunctionInput fi = new FunctionInput(name, funcs, Units.uAcm2);
      network.addNode(fi);
      
      //we can add a probe to it and run the simulator ... 
      Probe p = network.getSimulator().addProbe(name, FunctionInput.STATE_NAME, true);      
      network.run(0, 10);
      
      JFrame frame = makeAnimPlotFrame("vector plot");
      frame.setLayout(new BorderLayout());
      Scope scope = new Scope(p);
      frame.getContentPane().add(scope.getGraphPanel(), BorderLayout.CENTER);

View Full Code Here

    Plotter.plot(series, patterns, "multi series");
  }
  
  public static void rasterMemoryTest() {
    try {
      Network network = new NetworkImpl();
      
      NEFEnsembleFactory ef = new NEFEnsembleFactoryImpl();
      final NEFEnsemble ensemble = ef.make("ensemble", 200, 1);
      ensemble.collectSpikes(true);
      network.addNode(ensemble);
      
      FunctionInput input = new FunctionInput("input", new Function[]{new ConstantFunction(1, 1)}, Units.UNK);
      network.addNode(input);
      
      network.run(0, 1);
      
      Environment.setUserInterface(true);
      Memory.report("Before plots");
      for (int i = 0; i < 5; i++) {
        Thread pt = new Thread() {

View Full Code Here

      }
    };
    FunctionInput input = new FunctionInput("input", new Function[]{f}, Units.UNK);
//    FunctionInput zero = new FunctionInput("zero", new Function[]{new ConstantFunction(1, 0f)}, Units.UNK);
    
    Network network = new NetworkImpl();
    network.addNode(input);
    network.addNode(source);
    network.addNode(dest);


    source.addDecodedTermination("input", MU.I(1), .005f, false); //OK
    BiasOrigin bo = source.addBiasOrigin(source.getOrigin(NEFEnsemble.X), 200, "interneurons", true); //should have -ve bias decoders
    network.addNode(bo.getInterneurons()); //should be backwards response functions
//**    bo.getInterneurons().addDecodedTermination("source", MU.I(1), .005f, false);
    
//    Plotter.plot(bo.getInterneurons());
//    Plotter.plot(bo.getInterneurons(), NEFEnsemble.X);
    
//    DecodedTermination t = (DecodedTermination) dest.addDecodedTermination("source", MU.I(1), .005f, false);
//**    BiasTermination[] bt = dest.addBiasTerminations(t, .002f, bo.getDecoders()[0][0], ((DecodedOrigin) source.getOrigin(NEFEnsemble.X)).getDecoders());
//**    bt[1].setStaticBias(-1); //creates intrinsic current needed to counteract interneuron activity at 0
    
//    float[][] weights = MU.prod(dest.getEncoders(), MU.transpose(((DecodedOrigin) source.getOrigin(NEFEnsemble.X)).getDecoders()));
//*    float[][] biasEncoders = MU.transpose(new float[][]{bt[0].getBiasEncoders()});
//*    float[][] biasDecoders = MU.transpose(bo.getDecoders());
//*    float[][] weightBiases = MU.prod(biasEncoders, biasDecoders);
//*    float[][] biasedWeights = MU.sum(weights, weightBiases);
//    Plotter.plot(weights[0], "some weights");
//    Plotter.plot(biasedWeights[0], "some biased weights");
//    Plotter.plot(weights[1], "some more weights");
    
//    Plotter.plot(bt[0].getBiasEncoders(), "bias decoders");
    
    network.addProjection(input.getOrigin(FunctionInput.ORIGIN_NAME), source.getTermination("input"));
    network.addProjection(source.getOrigin(NEFEnsemble.X), dest.getTermination("source"));
//*    network.addProjection(bo, bo.getInterneurons().getTermination("source"));
//*    network.addProjection(bo, bt[0]);
//*    network.addProjection(bo.getInterneurons().getOrigin(NEFEnsemble.X), bt[1]);
//    network.addProjection(zero.getOrigin(FunctionInput.ORIGIN_NAME), bt[1]);
    
//    Probe sourceProbe = network.getSimulator().addProbe("source", NEFEnsemble.X, true);
//    Probe destProbe = network.getSimulator().addProbe("dest", NEFEnsemble.X, true);
//    Probe interProbe = network.getSimulator().addProbe("source_X_bias_interneurons", NEFEnsemble.X, true);
    
    network.run(0, 2);
    
//    Plotter.plot(sourceProbe.getData(), "source");
//    Plotter.plot(destProbe.getData(), "dest");
//    Plotter.plot(interProbe.getData(), "interneurons");
  }

View Full Code Here

  }
  
  public void functionalTestBiasOriginError() throws StructuralException, SimulationException {
    float tauPSC = .01f;
    
    Network network = new NetworkImpl();
    
    Function f = new AbstractFunction(1) {
      private static final long serialVersionUID = 1L;
      public float map(float[] from) {
        return from[0] - 1;
      }
    };
    
    FunctionInput input = new FunctionInput("input", new Function[]{f}, Units.UNK);
    network.addNode(input);
    
    NEFEnsembleFactory ef = new NEFEnsembleFactoryImpl();
    NEFEnsemble pre = ef.make("pre", 400, 1, "nefe_pre", false);
    pre.addDecodedTermination("input", MU.I(1), tauPSC, false);
//    DecodedOrigin baseOrigin = (DecodedOrigin) pre.getOrigin(NEFEnsemble.X);
    network.addNode(pre);
    
    NEFEnsemble post = ef.make("post", 200, 1, "nefe_post", false);
//    DecodedTermination baseTermination = (DecodedTermination) post.addDecodedTermination("pre", MU.I(1), tauPSC, false);
    network.addNode(post);
    
    network.addProjection(input.getOrigin(FunctionInput.ORIGIN_NAME), pre.getTermination("input"));
    Projection projection = network.addProjection(pre.getOrigin(NEFEnsemble.X), post.getTermination("pre"));
    
    Probe pPost = network.getSimulator().addProbe("post", NEFEnsemble.X, true);
    network.run(0, 2);
    TimeSeries ideal = pPost.getData();
    Plotter.plot(pPost.getData(), .005f, "mixed weights result");    
    
    //remove negative weights ... 
    System.out.println("Minimum weight without bias: " + MU.min(projection.getWeights()));
    projection.addBias(100, .005f, tauPSC, true, false);
    System.out.println("Minimum weight with bias: " + MU.min(projection.getWeights()));
    pPost.reset();
    network.run(0, 2);
    TimeSeries diff = new TimeSeriesImpl(ideal.getTimes(), MU.difference(ideal.getValues(), pPost.getData().getValues()), ideal.getUnits()); 
    Plotter.plot(diff, .01f, "positive weights");
    
    projection.removeBias();
    projection.addBias(100, tauPSC/5f, tauPSC, true, true);
    pPost.reset();
    Probe pInter = network.getSimulator().addProbe("post:pre:interneurons", NEFEnsemble.X, true);
    network.run(0, 2);
    diff = new TimeSeriesImpl(ideal.getTimes(), MU.difference(ideal.getValues(), pPost.getData().getValues()), ideal.getUnits()); 
    Plotter.plot(diff, .01f, "positive weights optimized");
    Plotter.plot(pInter.getData(), .01f, "interneurons");

View Full Code Here

          for (int i = 0; i < result.myNodeTerminations.length; i++){
            result.myNodeTerminations[i] = ensemble.getNodes()[i].getTermination(myNodeTerminations[i].getName());
          }
        }
        if (node instanceof Network) {
          Network network = (Network)node;
          for (int i = 0; i < result.myNodeTerminations.length; i++){
            result.myNodeTerminations[i] = network.getNodes()[i].getTermination(myNodeTerminations[i].getName());
          }
        }
      } catch (StructuralException e) {
        throw new CloneNotSupportedException("Error cloning EnsembleTermination: " + e.getMessage());
      }

View Full Code Here


    //functional test
    public static void main2(String[] args) {


        try {
            Network network = new NetworkImpl();


            //x, .3: varying x keeps time constant, changes adapted rate
//          ALIFSpikeGenerator generator = new ALIFSpikeGenerator(.002f, .02f, .5f, .01f);  //.2: .01 to .3 (150 to 20ms)
//          SynapticIntegrator integrator = new LinearSynapticIntegrator(.001f, Units.ACU);
//          PlasticExpandableSpikingNeuron neuron = new PlasticExpandableSpikingNeuron(integrator, generator, 15f, 0f, "alif");


            ALIFNeuronFactory factory = new ALIFNeuronFactory(new IndicatorPDF(200, 400), new IndicatorPDF(-2.5f, -1.5f),
                    new IndicatorPDF(.1f, .1001f), .0005f, .02f, .2f);


//          VectorGenerator vg = new RandomHypersphereVG(false, 1, 0);
//          ApproximatorFactory factory = new WeightedCostApproximator.Factory(.1f);
//          NEFEnsemble ensemble = new NEFEnsembleImpl("ensemble", new NEFNode[]{neuron}, new float[][]{new float[]{1}}, factory, vg.genVectors(100, 1));


            Node[] neurons = new Node[50];
            float[][] weights = new float[neurons.length][];
            for (int i = 0; i < neurons.length; i++) {
                neurons[i] = factory.make("neuron"+i);
                weights[i] = new float[]{1};
            }
            EnsembleImpl ensemble = new EnsembleImpl("ensemble", neurons);
            ensemble.addTermination("input", weights, .005f, false);
            ensemble.collectSpikes(true);
            network.addNode(ensemble);


            FunctionInput input = new FunctionInput("input", new Function[]{new PiecewiseConstantFunction(new float[]{0.2f}, new float[]{0, 0.5f})}, Units.UNK);
            network.addNode(input);


            network.addProjection(input.getOrigin(FunctionInput.ORIGIN_NAME), ensemble.getTermination("input"));


//          Probe vProbe = network.getSimulator().addProbe("ensemble", 0, "V", true);
//          Probe nProbe = network.getSimulator().addProbe("ensemble", 0, "N", true);
//          Probe iProbe = network.getSimulator().addProbe("ensemble", 0, "I", true);
            Probe rProbe = network.getSimulator().addProbe("ensemble", "rate", true);


            network.setMode(SimulationMode.RATE);
            network.run(0, 1);


//          Plotter.plot(ensemble.getSpikePattern());
//          Plotter.plot(vProbe.getData(), "V");
//          Plotter.plot(nProbe.getData(), "N");
//          Plotter.plot(iProbe.getData(), "I");

View Full Code Here

0 1 2

TOP

Related Classes of ca.nengo.model.Network

ca.nengo.examples.IntegratorExample

ca.nengo.model.impl.EnsembleOrigin

ca.nengo.model.impl.EnsembleTermination

ca.nengo.model.impl.NetworkImpl

ca.nengo.model.impl.ProjectionImplTest

ca.nengo.model.nef.impl.NEFEnsembleImplTest

ca.nengo.model.neuron.impl.ALIFSpikeGeneratorTest

ca.nengo.model.neuron.impl.IzhikevichSpikeGeneratorTest

ca.nengo.model.neuron.impl.PoissonSpikeGeneratorTest

ca.nengo.plot.PlotterFunctional

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.