neuroph简单的感知器测试

最新推荐文章于 2021-02-20 21:39:15 发布

iteye_16125

最新推荐文章于 2021-02-20 21:39:15 发布

阅读量158

点赞数

分类专栏：各种奇怪算法

本文链接：https://blog.csdn.net/iteye_16125/article/details/81864661

版权

各种奇怪算法专栏收录该内容

2 篇文章 0 订阅

订阅专栏

这是一个感知器学习AND规则的例子，抄自NEU的Sample

首先定义TrainingSet，两个参数分别表示输入向量SIZE和输出向量SIZE，看代码：


 /**
         * Creates an instance of new empty training set
         * 
         * @param inputVectorSize
         * @param outputVectorSize
         */
        public TrainingSet(int inputVectorSize, int outputVectorSize) {
		this.elements = new Vector<TrainingElement>();
                this.inputVectorSize = inputVectorSize;
                this.outputVectorSize = outputVectorSize;
	}

然后定义一个感知器，两个参数分别表示输入层的元数和输出的元数，代码：


/**
	 * Creates new Perceptron with specified number of neurons in input and
	 * output layer, with Step trqansfer function
	 * 
	 * @param inputNeuronsCount
	 *            number of neurons in input layer
	 * @param outputNeuronsCount
	 *            number of neurons in output layer
	 */
	public Perceptron(int inputNeuronsCount, int outputNeuronsCount) {
		this.createNetwork(inputNeuronsCount, outputNeuronsCount, TransferFunctionType.STEP);
	}

然后在当前线程启动学习过程。对于需要异步执行的，提供了learnInNewThread来满足需求。
训练网络

验证输出。



public void testPerceptron() {
        TrainingSet trainingSet = new TrainingSet(2, 1);
        trainingSet.addElement(new SupervisedTrainingElement(new double[]{0, 0}, new double[]{0}));
        trainingSet.addElement(new SupervisedTrainingElement(new double[]{0, 1}, new double[]{0}));
        trainingSet.addElement(new SupervisedTrainingElement(new double[]{1, 0}, new double[]{0}));
        trainingSet.addElement(new SupervisedTrainingElement(new double[]{1, 1}, new double[]{1}));

        NeuralNetwork myPerceptron = new Perceptron(2, 1);
        myPerceptron.learnInSameThread(trainingSet);
        System.out.println("Testing trained perceptron");
        neuralNetwork(myPerceptron, trainingSet);
        myPerceptron.save("mySamplePerceptron.nnet");


        NeuralNetwork loadedPerceptron = NeuralNetwork.load("mySamplePerceptron.nnet");
        System.out.println("Testing loaded perceptron");
        neuralNetwork(loadedPerceptron, trainingSet);
	}



	private void neuralNetwork(NeuralNetwork neuralNet, TrainingSet trainingSet) {

	    for(TrainingElement trainingElement : trainingSet.trainingElements()) {
	        neuralNet.setInput(trainingElement.getInput());
	        neuralNet.calculate();
	        Vector<Double> networkOutput = neuralNet.getOutput();

	        double d1 = trainingElement.getInput().get(0);
	        double d2 = trainingElement.getInput().get(1);

	        double result = networkOutput.get(0);

	        boolean b1 = d1 == 1.0;
	        boolean b2 = d2 == 1.0;

	        boolean r1 = result == 1.0;

	        assertEquals(b1 & b2, r1);
	    }
	}