backpropagationpanel.java

利用Java实现的神经网络工具箱

		errorPanel = new JPanel();
		errorPanel.setLayout(new BorderLayout());
		errorPanel.add(errorLabel, BorderLayout.NORTH);
		errorPanel.add(errorSpinner, BorderLayout.CENTER);
		
		stopConditionPanel = new JPanel();
		stopConditionPanel.setLayout(new GridBagLayout());
		stopConditionPanel.add(maxEpochsPanel,
				new GridBagConstraints(0, 0, 1, 1, 1.0, 1.0,
						GridBagConstraints.CENTER,
						GridBagConstraints.VERTICAL, borderInsets,
						67, 0));
		stopConditionPanel.add(errorPanel,
				new GridBagConstraints(0, 1, 1, 1, 1.0, 1.0,
						GridBagConstraints.CENTER,
						GridBagConstraints.VERTICAL, borderInsets,
						67, 0));
		stopConditionPanel.setBorder(new TitledBorder(resource
				.getString("stopCondition")));
		
		lrMultiplierLabel = new JLabel(resource.getString("lrMultiplier"));
		lrMultiplierSpinnerModel = new SpinnerNumberModel(10, 1e-30,
				1e10, 0.001);
		lrMultiplierSpinner = new JSpinner(lrMultiplierSpinnerModel);
		lrMultiplierPanel = new JPanel();
		lrMultiplierPanel.setLayout(new BorderLayout());
		lrMultiplierPanel.add(lrMultiplierLabel, BorderLayout.NORTH);
		lrMultiplierPanel.add(lrMultiplierSpinner, BorderLayout.CENTER);
		
		learningRateLabel = new JLabel(resource.getString("learningRate"));
		learningRateSpinnerModel = new SpinnerNumberModel(0.3, 1e-30, 1, 0.001);
		learningRateSpinner = new JSpinner(learningRateSpinnerModel);
		learningRatePanel = new JPanel();
		learningRatePanel.setLayout(new BorderLayout());
		learningRatePanel.add(learningRateLabel, BorderLayout.NORTH);
		learningRatePanel.add(learningRateSpinner, BorderLayout.CENTER);
		
		alphaLabel = new JLabel(resource.getString("alpha"));
		alphaSpinnerModel = new SpinnerNumberModel(0.2, 1e-30, 1, 0.001);
		alphaSpinner = new JSpinner(alphaSpinnerModel);
		alphaPanel = new JPanel();
		alphaPanel.setLayout(new BorderLayout());
		alphaPanel.add(alphaLabel, BorderLayout.NORTH);
		alphaPanel.add(alphaSpinner, BorderLayout.CENTER);
		
		backPropagationParametersPanel = new JPanel();
		backPropagationParametersPanel.setLayout(new GridBagLayout());
		backPropagationParametersPanel.add(lrMultiplierPanel,
				new GridBagConstraints(0, 0, 1, 1, 1.0, 1.0,
						GridBagConstraints.CENTER,
						GridBagConstraints.VERTICAL, borderInsets,
						67, 0));
		
		backPropagationParametersPanel.add(learningRatePanel,
				new GridBagConstraints(0, 1, 1, 1, 1.0, 1.0,
						GridBagConstraints.CENTER,
						GridBagConstraints.VERTICAL, borderInsets,
						67, 0));
		backPropagationParametersPanel.add(alphaPanel,
				new GridBagConstraints(0, 2, 1, 1, 1.0, 1.0,
						GridBagConstraints.CENTER,
						GridBagConstraints.VERTICAL, borderInsets,
						67, 0));
		backPropagationParametersPanel.setBorder(new TitledBorder(resource
				.getString("backPropagationParameters")));
		
		this.setLayout(new GridBagLayout());
		this.add(backPropagationVariationPanel,
				new GridBagConstraints(0, 0, 1, 1, 1.0, 1.0,
						GridBagConstraints.CENTER,
						GridBagConstraints.HORIZONTAL, borderInsets,
						67, 0));
		this.add(stopConditionPanel,
				new GridBagConstraints(0, 1, 1, 1, 1.0, 1.0,
						GridBagConstraints.CENTER,
						GridBagConstraints.HORIZONTAL, borderInsets,
						67, 0));
		this.add(backPropagationParametersPanel,
				new GridBagConstraints(0, 2, 1, 1, 1.0, 1.0,
						GridBagConstraints.CENTER,
						GridBagConstraints.HORIZONTAL, borderInsets,
						67, 0));
		
	} //GaussNewtonPanel()

	/* 
	 * @see neuralnetworktoolkit.gui.options.method.TrainingMethodPanel#initializeComponents(boolean)
	 */
	public void initializeComponents(boolean verify) {
		BackPropagationParameters bpParameters;
		
		if(verify) {
			bpParameters = (BackPropagationParameters) parameters;
			
			if (bpParameters.getMethod().equals(methods[0])) {
				backPropagationVariationGroup.setSelected(
						batchRadio.getModel(), true);
				setBackpropagationParametersEnabled(LEARNING_RATE_ENABLED);

			} else if (bpParameters.getMethod().equals(methods[1])) {
				backPropagationVariationGroup.setSelected(
						batchAdaptiveLearningRateRadio.getModel(), true);
				setBackpropagationParametersEnabled(ALL_ENABLED);

			} else if (bpParameters.getMethod().equals(methods[2])) {
				backPropagationVariationGroup.setSelected(batchMomentumRadio
						.getModel(), true);
				setBackpropagationParametersEnabled(MULTIPLIER_DISABLED);

			} else if (bpParameters.getMethod().equals(methods[3])) {
				backPropagationVariationGroup.setSelected(onLineRadio
						.getModel(), true);
				setBackpropagationParametersEnabled(LEARNING_RATE_ENABLED);

			} else if (bpParameters.getMethod().equals(methods[4])) {
				backPropagationVariationGroup.setSelected(
						onLineAdaptiveLearningRadio.getModel(), true);
				setBackpropagationParametersEnabled(ALL_ENABLED);

			} else if (bpParameters.getMethod().equals(methods[5])) {
				backPropagationVariationGroup.setSelected(onLineMomentumRadio
						.getModel(), true);
				setBackpropagationParametersEnabled(MULTIPLIER_DISABLED);

			}
			errorSpinner.setValue(new Double(bpParameters.getError()));
			maxEpochsSpinner.setValue(new Integer(bpParameters.getMaxEpochs()));
			lrMultiplierSpinner.setValue(new Double(bpParameters
					.getLrMultiplier()));
			learningRateSpinner.setValue(new Double(bpParameters
					.getLearningRate()));
			alphaSpinner.setValue(new Double(bpParameters.getAlpha()));
			
		}
		
	} //initializeComponents()
	
	/**
	 * @return Returns the parameters.
	 */
	public TrainingParameters getParameters() {
		BackPropagationParameters bpParameters;
		
		bpParameters = (BackPropagationParameters) parameters;
		
		if(batchRadio.isSelected()) {
			bpParameters.setMethod(methods[0]);
			
		} else if(batchAdaptiveLearningRateRadio.isSelected()) {
			bpParameters.setMethod(methods[1]);
			
		} else if(batchMomentumRadio.isSelected()) {
			bpParameters.setMethod(methods[2]);
			
		} else if(onLineRadio.isSelected()) {
			bpParameters.setMethod(methods[3]);
			
		} else if(onLineAdaptiveLearningRadio.isSelected()) {
			bpParameters.setMethod(methods[4]);
			
		} else if(onLineMomentumRadio.isSelected()) {
			bpParameters.setMethod(methods[5]);
			
		}
		bpParameters.setError(((Double) errorSpinner.getValue()).doubleValue());
		bpParameters.setMaxEpochs(((Integer) maxEpochsSpinner.getValue())
				.intValue());
		bpParameters.setLearningRate(((Double) learningRateSpinner.getValue())
				.doubleValue());
		bpParameters.setAlpha(((Double) alphaSpinner.getValue()).doubleValue());
		bpParameters.setLrMultiplier(((Double) lrMultiplierSpinner.getValue())
				.doubleValue());
		
		return bpParameters;
		
	} //getParameters()
	
	/**
	 * 
	 * @param type
	 */
	private void setBackpropagationParametersEnabled(int type) {
		
		switch (type) {
		case ALL_ENABLED: {
			lrMultiplierSpinner.setEnabled(true);
			learningRateSpinner.setEnabled(true);
			alphaSpinner.setEnabled(true);
			
		} break;
		case LEARNING_RATE_ENABLED: {
			lrMultiplierSpinner.setEnabled(false);
			learningRateSpinner.setEnabled(true);
			alphaSpinner.setEnabled(false);
			
		} break;
		case MULTIPLIER_DISABLED: {
			lrMultiplierSpinner.setEnabled(false);
			learningRateSpinner.setEnabled(true);
			alphaSpinner.setEnabled(true);
			
		} break;
		default: {
			lrMultiplierSpinner.setEnabled(true);
			learningRateSpinner.setEnabled(true);
			alphaSpinner.setEnabled(true);
			
		} break;
		
		}
		
	} //setBackpropagationParametersEnabled()

} //BackPropagationPanel