Examples of Ensemble

ca.nengo.model.Ensemble
A group of Nodes with largely overlapping inputs and outputs.

There are no strict rules for how to group Nodes into Ensembles, but here are some things to consider:
- A group of Nodes that together 'represent' something through a population code should be modelled as an Ensemble. (Also consider using NEFEnsemble to make such representation explicit.)
- Making ensembles that correspond to physical structures (e.g. nuclei) and naming them appropriately will make the model clearer.
- Outputs from an Ensemble are grouped together and passed to other Ensembles during a simulation, and practical issues may arise from this. For example, putting all your Nodes in a single large ensemble could result in a very large matrix of synaptic weights, which would impair performance.
The membership of an Ensemble is fixed once the Ensemble is created. This means that the Ensemble model doesn't deal explicitly with growth and death of components during simulation (although you can set input/output weights to zero to mimic this). It also means that an Ensemble isn't a good model of a functional "assembly".
@author Bryan Tripp
edu.uci.jforestsx.learning.trees.Ensemble
@author Yasser Ganjisaffar

Examples of edu.uci.jforestsx.learning.trees.Ensemble

    curTrainSet = trainSet;
    curValidSet = validSet;


    preprocess();


    Ensemble ensemble = new Ensemble();
    bestValidationMeasurement = Double.NaN;
    int earlyStoppingIteration = 0;
    int bestIteration = 0;
    int[] treeCounts = new int[numSubModules];
    subLearner.setTreeWeight(treeWeight);
    for (curIteration = 1; curIteration <= numSubModules; curIteration++) {
      Sample subLeanerSample = getSubLearnerSample();
      Ensemble subEnsemble = subLearner.learn(subLeanerSample, validSet);
      if (subEnsemble == null) {
        break;
      }
      for (int t = 0; t < subEnsemble.getNumTrees(); t++) {
        Tree tree = subEnsemble.getTreeAt(t);
        ensemble.addTree(tree, subEnsemble.getWeightAt(t));
        if (validSet != null) {
          LearningUtils.updateScores(validSet, validPredictions, ((RegressionTree) tree), 1.0);
        }
      }
      treeCounts[curIteration - 1] = ensemble.getNumTrees();

View Full Code Here

Examples of edu.uci.jforestsx.learning.trees.Ensemble


    if (options.has("validation-file")) {
      configProperties.put(TrainingConfig.VALID_FILENAME, options.valueOf("validation-file"));
    }


    Ensemble ensemble;


    if (options.has("ranking")) {
      RankingApp app = new RankingApp();
      ensemble = app.run(configProperties);
    } else {

View Full Code Here

Examples of edu.uci.jforestsx.learning.trees.Ensemble


    /*
     * Load the ensemble
     */
    File modelFile = new File((String) options.valueOf("model-file"));
    Ensemble ensemble = new Ensemble();
    if (options.valueOf("tree-type").equals("RegressionTree")) {
      ensemble.loadFromFile(RegressionTree.class, modelFile);
    } else if (options.valueOf("tree-type").equals("DecisionTree")) {
      ensemble.loadFromFile(DecisionTree.class, modelFile);
    } else {
      System.err.println("Unknown tree type: " + options.valueOf("tree-type"));
    }


    /*

View Full Code Here

Examples of edu.uci.jforestsx.learning.trees.Ensemble


      createLearner();


      Timer timer = new Timer();
      timer.start();
      Ensemble ensemble = topLearner.learn(trainSet, validSet);
      System.out.println("Time taken to build model: " + (timer.getElapsedMillis() / 1000.0) + " seconds.");
      return ensemble;


    } catch (Exception e) {
      e.printStackTrace();

View Full Code Here

0 1

TOP

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.