nampdn · June 26, 2024 18:52
diff --git a/gistfile1.txt b/gistfile1.txt
 <!DOCTYPE html>
 <html lang="en">
 <head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <title>Neural Network Training on Sine Function</title>
    <script src="https://cdnjs.cloudflare.com/ajax/libs/Chart.js/3.7.0/chart.min.js"></script>
    <style>
        body { font-family: Arial, sans-serif; margin: 0; padding: 20px; display: flex; flex-direction: column; align-items: center; }
        #visualization-container { display: flex; justify-content: space-around; width: 100%; margin-bottom: 20px; flex-wrap: wrap; }
        canvas { border: 1px solid #ccc; margin: 10px; }
        #controls { margin-top: 20px; }
        button, input { margin: 5px; }
    </style>
 </head>
 <body>
    <h1>Neural Network Training on Sine Function</h1>
    <div id="visualization-container">
        <canvas id="networkCanvas" width="800" height="400"></canvas>
        <canvas id="functionPlot" width="400" height="300"></canvas>
        <canvas id="errorChart" width="400" height="300"></canvas>
    </div>
    <div id="controls">
        <button id="startBtn">Start Training</button>
        <button id="resetBtn">Reset</button>
        <div>
            Learning Rate: <span id="learningRateValue">0.01</span>
            <input type="range" id="learningRateSlider" min="0.001" max="0.1" step="0.001" value="0.01">
        </div>
    </div>

    <script>
        const networkCanvas = document.getElementById('networkCanvas');
        const networkCtx = networkCanvas.getContext('2d');
        const startBtn = document.getElementById('startBtn');
        const resetBtn = document.getElementById('resetBtn');
        const learningRateSlider = document.getElementById('learningRateSlider');
        const learningRateValue = document.getElementById('learningRateValue');

        // Neural Network parameters
        const layers = [1, 10, 10, 1];  // Input, Hidden layers, Output
        let weights = [];
        let biases = [];
        let activations = [];
        let learningRate = 0.01;
        let isTraining = false;
        let epoch = 0;
        const maxEpochs = 10000;

        // Initialize weights, biases, and activations
        function initializeNetwork() {
            weights = [];
            biases = [];
            activations = [new Array(layers[0]).fill(0)];
            for (let i = 1; i < layers.length; i++) {
                const layerWeights = [];
                const layerBiases = [];
                for (let j = 0; j < layers[i]; j++) {
                    const neuronWeights = [];
                    for (let k = 0; k < layers[i-1]; k++) {
                        neuronWeights.push(Math.random() - 0.5);
                    }
                    layerWeights.push(neuronWeights);
                    layerBiases.push(Math.random() - 0.5);
                }
                weights.push(layerWeights);
                biases.push(layerBiases);
                activations.push(new Array(layers[i]).fill(0));
            }
        }

        // Activation function (tanh)
        function tanh(x) {
            return Math.tanh(x);
        }

        // Forward pass
        function forwardPass(input) {
            activations[0] = [input];
            for (let i = 1; i < layers.length; i++) {
                for (let j = 0; j < layers[i]; j++) {
                    let sum = biases[i-1][j];
                    for (let k = 0; k < layers[i-1]; k++) {
                        sum += weights[i-1][j][k] * activations[i-1][k];
                    }
                    activations[i][j] = tanh(sum);
                }
            }
            return activations[activations.length - 1][0];
        }

        // Backpropagation
        function backpropagate(input, target) {
            const output = forwardPass(input);
            const outputError = output - target;
            
            // Calculate gradients and update weights/biases
            const gradients = [];
            for (let i = layers.length - 1; i > 0; i--) {
                const layerGradients = [];
                for (let j = 0; j < layers[i]; j++) {
                    let gradient;
                    if (i === layers.length - 1) {
                        gradient = outputError * (1 - Math.pow(activations[i][j], 2));
                    } else {
                        gradient = 0;
                        for (let k = 0; k < layers[i+1]; k++) {
                            gradient += gradients[0][k] * weights[i][k][j];
                        }
                        gradient *= (1 - Math.pow(activations[i][j], 2));
                    }
                    layerGradients.push(gradient);
                    
                    for (let k = 0; k < layers[i-1]; k++) {
                        weights[i-1][j][k] -= learningRate * gradient * activations[i-1][k];
                    }
                    biases[i-1][j] -= learningRate * gradient;
                }
                gradients.unshift(layerGradients);
            }

            return outputError ** 2;
        }

        // Training step
        function trainingStep() {
            const input = Math.random() * 2 * Math.PI;
            const target = Math.sin(input);
            return backpropagate(input, target);
        }

        // Draw neural network
        function drawNetwork() {
            networkCtx.clearRect(0, 0, networkCanvas.width, networkCanvas.height);
            const layerWidth = networkCanvas.width / (layers.length + 1);
            const maxNeurons = Math.max(...layers);
            
            for (let i = 0; i < layers.length; i++) {
                const neurons = layers[i];
                for (let j = 0; j < neurons; j++) {
                    const x = (i + 1) * layerWidth;
                    const y = (j + 1) * (networkCanvas.height / (neurons + 1));
                    
                    // Draw neuron
                    networkCtx.beginPath();
                    networkCtx.arc(x, y, 15, 0, 2 * Math.PI);
                    networkCtx.fillStyle = 'lightblue';
                    networkCtx.fill();
                    networkCtx.stroke();
                    
                    // Draw activation value
                    networkCtx.fillStyle = 'black';
                    networkCtx.font = '10px Arial';
                    networkCtx.textAlign = 'center';
                    networkCtx.fillText(activations[i][j].toFixed(2), x, y);

                    // Draw bias (except for input layer)
                    if (i > 0) {
                        networkCtx.fillText('b: ' + biases[i-1][j].toFixed(2), x, y + 25);
                    }
                    
                    // Draw connections to next layer
                    if (i < layers.length - 1) {
                        const nextNeurons = layers[i + 1];
                        for (let k = 0; k < nextNeurons; k++) {
                            const nextX = (i + 2) * layerWidth;
                            const nextY = (k + 1) * (networkCanvas.height / (nextNeurons + 1));
                            const weight = weights[i][k][j];
                            const normalizedWeight = (weight + 1) / 2; // Normalize to [0, 1]
                            
                            // Draw connection
                            networkCtx.beginPath();
                            networkCtx.moveTo(x + 15, y);
                            networkCtx.lineTo(nextX - 15, nextY);
                            networkCtx.strokeStyle = `rgb(${255 * (1 - normalizedWeight)}, 0, ${255 * normalizedWeight})`;
                            networkCtx.lineWidth = Math.abs(weight) * 2;
                            networkCtx.stroke();
                            
                            // Draw "zap" effect
                            if (isTraining && Math.random() < 0.1) {
                                networkCtx.beginPath();
                                networkCtx.moveTo(x + 15, y);
                                for (let t = 0.1; t < 1; t += 0.1) {
                                    const zapX = x + 15 + (nextX - x - 30) * t;
                                    const zapY = y + (nextY - y) * t + (Math.random() - 0.5) * 10;
                                    networkCtx.lineTo(zapX, zapY);
                                }
                                networkCtx.lineTo(nextX - 15, nextY);
                                networkCtx.strokeStyle = 'yellow';
                                networkCtx.lineWidth = 2;
                                networkCtx.stroke();
                            }
                        }
                    }
                }
            }
        }

        // Function plot
        const functionPlot = new Chart(document.getElementById('functionPlot').getContext('2d'), {
            type: 'scatter',
            data: {
                datasets: [{
                    label: 'Target (sin)',
                    data: [],
                    borderColor: 'rgb(75, 192, 192)',
                    showLine: true,
                    pointRadius: 0
                }, {
                    label: 'Network Output',
                    data: [],
                    borderColor: 'rgb(255, 99, 132)',
                    showLine: true,
                    pointRadius: 0
                }]
            },
            options: {
                responsive: true,
                scales: {
                    x: {
                        type: 'linear',
                        position: 'bottom',
                        min: 0,
                        max: 2 * Math.PI
                    },
                    y: {
                        min: -1,
                        max: 1
                    }
                }
            }
        });

        // Update function plot
        function updateFunctionPlot() {
            const points = 100;
            const targetData = [];
            const outputData = [];
            for (let i = 0; i <= points; i++) {
                const x = (i / points) * 2 * Math.PI;
                targetData.push({x: x, y: Math.sin(x)});
                outputData.push({x: x, y: forwardPass(x)});
            }
            functionPlot.data.datasets[0].data = targetData;
            functionPlot.data.datasets[1].data = outputData;
            functionPlot.update();
        }

        // Error chart
        const errorChart = new Chart(document.getElementById('errorChart').getContext('2d'), {
            type: 'line',
            data: {
                labels: [],
                datasets: [{
                    label: 'Mean Squared Error',
                    data: [],
                    borderColor: 'rgb(75, 192, 192)',
                    tension: 0.1
                }]
            },
            options: {
                responsive: true,
                scales: {
                    y: {
                        beginAtZero: true
                    }
                },
                animation: {
                    duration: 0
                }
            }
        });

        // Update error chart
        function updateErrorChart(error) {
            errorChart.data.labels.push(epoch);
            errorChart.data.datasets[0].data.push(error);
            if (errorChart.data.labels.length > 100) {
                errorChart.data.labels.shift();
                errorChart.data.datasets[0].data.shift();
            }
            errorChart.update();
        }

        // Training loop
        function train() {
            if (isTraining && epoch < maxEpochs) {
                const error = trainingStep();
                if (epoch % 10 === 0) {
                    drawNetwork();
                    updateFunctionPlot();
                    updateErrorChart(error);
                }
                epoch++;
                requestAnimationFrame(train);
            } else {
                isTraining = false;
                startBtn.textContent = 'Start Training';
            }
        }

        // Reset network
        function resetNetwork() {
            initializeNetwork();
            epoch = 0;
            errorChart.data.labels = [];
            errorChart.data.datasets[0].data = [];
            errorChart.update();
            drawNetwork();
            updateFunctionPlot();
        }

        // Event listeners
        startBtn.addEventListener('click', () => {
            isTraining = !isTraining;
            startBtn.textContent = isTraining ? 'Pause Training' : 'Resume Training';
            if (isTraining) train();
        });

        resetBtn.addEventListener('click', resetNetwork);

        learningRateSlider.addEventListener('input', (e) => {
            learningRate = parseFloat(e.target.value);
            learningRateValue.textContent = learningRate.toFixed(3);
        });

        // Initialize
        initializeNetwork();
        drawNetwork();
        updateFunctionPlot();
    </script>
 </body>
 </html>
	<!DOCTYPE html>
	<html lang="en">
	<head>
	<meta charset="UTF-8">
	<meta name="viewport" content="width=device-width, initial-scale=1.0">
	<title>Neural Network Training on Sine Function</title>
	<script src="https://cdnjs.cloudflare.com/ajax/libs/Chart.js/3.7.0/chart.min.js"></script>
	<style>
	body { font-family: Arial, sans-serif; margin: 0; padding: 20px; display: flex; flex-direction: column; align-items: center; }
	#visualization-container { display: flex; justify-content: space-around; width: 100%; margin-bottom: 20px; flex-wrap: wrap; }
	canvas { border: 1px solid #ccc; margin: 10px; }
	#controls { margin-top: 20px; }
	button, input { margin: 5px; }
	</style>
	</head>
	<body>
	<h1>Neural Network Training on Sine Function</h1>
	<div id="visualization-container">
	<canvas id="networkCanvas" width="800" height="400"></canvas>
	<canvas id="functionPlot" width="400" height="300"></canvas>
	<canvas id="errorChart" width="400" height="300"></canvas>
	</div>
	<div id="controls">
	<button id="startBtn">Start Training</button>
	<button id="resetBtn">Reset</button>
	<div>
	Learning Rate: <span id="learningRateValue">0.01</span>
	<input type="range" id="learningRateSlider" min="0.001" max="0.1" step="0.001" value="0.01">
	</div>
	</div>

	<script>
	const networkCanvas = document.getElementById('networkCanvas');
	const networkCtx = networkCanvas.getContext('2d');
	const startBtn = document.getElementById('startBtn');
	const resetBtn = document.getElementById('resetBtn');
	const learningRateSlider = document.getElementById('learningRateSlider');
	const learningRateValue = document.getElementById('learningRateValue');

	// Neural Network parameters
	const layers = [1, 10, 10, 1]; // Input, Hidden layers, Output
	let weights = [];
	let biases = [];
	let activations = [];
	let learningRate = 0.01;
	let isTraining = false;
	let epoch = 0;
	const maxEpochs = 10000;

	// Initialize weights, biases, and activations
	function initializeNetwork() {
	weights = [];
	biases = [];
	activations = [new Array(layers[0]).fill(0)];
	for (let i = 1; i < layers.length; i++) {
	const layerWeights = [];
	const layerBiases = [];
	for (let j = 0; j < layers[i]; j++) {
	const neuronWeights = [];
	for (let k = 0; k < layers[i-1]; k++) {
	neuronWeights.push(Math.random() - 0.5);
	}
	layerWeights.push(neuronWeights);
	layerBiases.push(Math.random() - 0.5);
	}
	weights.push(layerWeights);
	biases.push(layerBiases);
	activations.push(new Array(layers[i]).fill(0));
	}
	}

	// Activation function (tanh)
	function tanh(x) {
	return Math.tanh(x);
	}

	// Forward pass
	function forwardPass(input) {
	activations[0] = [input];
	for (let i = 1; i < layers.length; i++) {
	for (let j = 0; j < layers[i]; j++) {
	let sum = biases[i-1][j];
	for (let k = 0; k < layers[i-1]; k++) {
	sum += weights[i-1][j][k] * activations[i-1][k];
	}
	activations[i][j] = tanh(sum);
	}
	}
	return activations[activations.length - 1][0];
	}

	// Backpropagation
	function backpropagate(input, target) {
	const output = forwardPass(input);
	const outputError = output - target;

	// Calculate gradients and update weights/biases
	const gradients = [];
	for (let i = layers.length - 1; i > 0; i--) {
	const layerGradients = [];
	for (let j = 0; j < layers[i]; j++) {
	let gradient;
	if (i === layers.length - 1) {
	gradient = outputError * (1 - Math.pow(activations[i][j], 2));
	} else {
	gradient = 0;
	for (let k = 0; k < layers[i+1]; k++) {
	gradient += gradients[0][k] * weights[i][k][j];
	}
	gradient *= (1 - Math.pow(activations[i][j], 2));
	}
	layerGradients.push(gradient);

	for (let k = 0; k < layers[i-1]; k++) {
	weights[i-1][j][k] -= learningRate * gradient * activations[i-1][k];
	}
	biases[i-1][j] -= learningRate * gradient;
	}
	gradients.unshift(layerGradients);
	}

	return outputError ** 2;
	}

	// Training step
	function trainingStep() {
	const input = Math.random() * 2 * Math.PI;
	const target = Math.sin(input);
	return backpropagate(input, target);
	}

	// Draw neural network
	function drawNetwork() {
	networkCtx.clearRect(0, 0, networkCanvas.width, networkCanvas.height);
	const layerWidth = networkCanvas.width / (layers.length + 1);
	const maxNeurons = Math.max(...layers);

	for (let i = 0; i < layers.length; i++) {
	const neurons = layers[i];
	for (let j = 0; j < neurons; j++) {
	const x = (i + 1) * layerWidth;
	const y = (j + 1) * (networkCanvas.height / (neurons + 1));

	// Draw neuron
	networkCtx.beginPath();
	networkCtx.arc(x, y, 15, 0, 2 * Math.PI);
	networkCtx.fillStyle = 'lightblue';
	networkCtx.fill();
	networkCtx.stroke();

	// Draw activation value
	networkCtx.fillStyle = 'black';
	networkCtx.font = '10px Arial';
	networkCtx.textAlign = 'center';
	networkCtx.fillText(activations[i][j].toFixed(2), x, y);

	// Draw bias (except for input layer)
	if (i > 0) {
	networkCtx.fillText('b: ' + biases[i-1][j].toFixed(2), x, y + 25);
	}

	// Draw connections to next layer
	if (i < layers.length - 1) {
	const nextNeurons = layers[i + 1];
	for (let k = 0; k < nextNeurons; k++) {
	const nextX = (i + 2) * layerWidth;
	const nextY = (k + 1) * (networkCanvas.height / (nextNeurons + 1));
	const weight = weights[i][k][j];
	const normalizedWeight = (weight + 1) / 2; // Normalize to [0, 1]

	// Draw connection
	networkCtx.beginPath();
	networkCtx.moveTo(x + 15, y);
	networkCtx.lineTo(nextX - 15, nextY);
	networkCtx.strokeStyle = `rgb(${255 * (1 - normalizedWeight)}, 0, ${255 * normalizedWeight})`;
	networkCtx.lineWidth = Math.abs(weight) * 2;
	networkCtx.stroke();

	// Draw "zap" effect
	if (isTraining && Math.random() < 0.1) {
	networkCtx.beginPath();
	networkCtx.moveTo(x + 15, y);
	for (let t = 0.1; t < 1; t += 0.1) {
	const zapX = x + 15 + (nextX - x - 30) * t;
	const zapY = y + (nextY - y) * t + (Math.random() - 0.5) * 10;
	networkCtx.lineTo(zapX, zapY);
	}
	networkCtx.lineTo(nextX - 15, nextY);
	networkCtx.strokeStyle = 'yellow';
	networkCtx.lineWidth = 2;
	networkCtx.stroke();
	}
	}
	}
	}
	}
	}

	// Function plot
	const functionPlot = new Chart(document.getElementById('functionPlot').getContext('2d'), {
	type: 'scatter',
	data: {
	datasets: [{
	label: 'Target (sin)',
	data: [],
	borderColor: 'rgb(75, 192, 192)',
	showLine: true,
	pointRadius: 0
	}, {
	label: 'Network Output',
	data: [],
	borderColor: 'rgb(255, 99, 132)',
	showLine: true,
	pointRadius: 0
	}]
	},
	options: {
	responsive: true,
	scales: {
	x: {
	type: 'linear',
	position: 'bottom',
	min: 0,
	max: 2 * Math.PI
	},
	y: {
	min: -1,
	max: 1
	}
	}
	}
	});

	// Update function plot
	function updateFunctionPlot() {
	const points = 100;
	const targetData = [];
	const outputData = [];
	for (let i = 0; i <= points; i++) {
	const x = (i / points) * 2 * Math.PI;
	targetData.push({x: x, y: Math.sin(x)});
	outputData.push({x: x, y: forwardPass(x)});
	}
	functionPlot.data.datasets[0].data = targetData;
	functionPlot.data.datasets[1].data = outputData;
	functionPlot.update();
	}

	// Error chart
	const errorChart = new Chart(document.getElementById('errorChart').getContext('2d'), {
	type: 'line',
	data: {
	labels: [],
	datasets: [{
	label: 'Mean Squared Error',
	data: [],
	borderColor: 'rgb(75, 192, 192)',
	tension: 0.1
	}]
	},
	options: {
	responsive: true,
	scales: {
	y: {
	beginAtZero: true
	}
	},
	animation: {
	duration: 0
	}
	}
	});

	// Update error chart
	function updateErrorChart(error) {
	errorChart.data.labels.push(epoch);
	errorChart.data.datasets[0].data.push(error);
	if (errorChart.data.labels.length > 100) {
	errorChart.data.labels.shift();
	errorChart.data.datasets[0].data.shift();
	}
	errorChart.update();
	}

	// Training loop
	function train() {
	if (isTraining && epoch < maxEpochs) {
	const error = trainingStep();
	if (epoch % 10 === 0) {
	drawNetwork();
	updateFunctionPlot();
	updateErrorChart(error);
	}
	epoch++;
	requestAnimationFrame(train);
	} else {
	isTraining = false;
	startBtn.textContent = 'Start Training';
	}
	}

	// Reset network
	function resetNetwork() {
	initializeNetwork();
	epoch = 0;
	errorChart.data.labels = [];
	errorChart.data.datasets[0].data = [];
	errorChart.update();
	drawNetwork();
	updateFunctionPlot();
	}

	// Event listeners
	startBtn.addEventListener('click', () => {
	isTraining = !isTraining;
	startBtn.textContent = isTraining ? 'Pause Training' : 'Resume Training';
	if (isTraining) train();
	});

	resetBtn.addEventListener('click', resetNetwork);

	learningRateSlider.addEventListener('input', (e) => {
	learningRate = parseFloat(e.target.value);
	learningRateValue.textContent = learningRate.toFixed(3);
	});

	// Initialize
	initializeNetwork();
	drawNetwork();
	updateFunctionPlot();
	</script>
	</body>
	</html>