File size: 22,935 Bytes
a895648 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 |
// Backpropagation Animation and Tab Functionality
document.addEventListener('DOMContentLoaded', () => {
// Set initialization flag
window.backpropInitialized = true;
console.log('Backpropagation script initialized');
// Canvas initialization function
function initializeCanvas() {
console.log('Initializing backpropagation canvas');
const canvas = document.getElementById('backprop-canvas');
if (!canvas) {
console.error('Backpropagation canvas not found!');
return;
}
const ctx = canvas.getContext('2d');
if (!ctx) {
console.error('Could not get 2D context for backpropagation canvas');
return;
}
// Set canvas dimensions
const container = canvas.parentElement;
if (container) {
canvas.width = container.clientWidth || 800;
canvas.height = container.clientHeight || 400;
} else {
canvas.width = 800;
canvas.height = 400;
}
// Clear canvas
ctx.clearRect(0, 0, canvas.width, canvas.height);
// Reset animation state and redraw
resetAnimation();
drawNetwork();
}
// Register the canvas initialization function with tab manager
if (typeof window !== 'undefined') {
window.initBackpropCanvas = initializeCanvas;
}
// Tab functionality
const tabButtons = document.querySelectorAll('.tab-button');
const tabContents = document.querySelectorAll('.tab-content');
tabButtons.forEach(button => {
button.addEventListener('click', () => {
// Remove active class from all tabs
tabButtons.forEach(btn => btn.classList.remove('active'));
tabContents.forEach(content => content.classList.remove('active'));
// Add active class to clicked tab
button.classList.add('active');
const tabId = button.getAttribute('data-tab');
document.getElementById(`${tabId}-tab`).classList.add('active');
// If switching to backpropagation tab, reset the animation
if (tabId === 'backpropagation') {
resetAnimation();
}
});
});
// Backpropagation Animation Setup
const canvas = document.getElementById('backprop-canvas');
const ctx = canvas.getContext('2d');
// Animation control buttons
const startButton = document.getElementById('start-animation');
const pauseButton = document.getElementById('pause-animation');
const resetButton = document.getElementById('reset-animation');
const speedControl = document.getElementById('animation-speed');
// Animation state
let animationState = {
running: false,
currentStep: 0,
speed: 5,
animationFrameId: null,
network: null,
lastTimestamp: 0
};
// Sample neural network for demonstration
class NeuralNetwork {
constructor() {
// Simple network with input, hidden and output layers
this.layers = [
{ type: 'input', neurons: 3, activation: 'none' },
{ type: 'hidden', neurons: 4, activation: 'relu' },
{ type: 'output', neurons: 2, activation: 'sigmoid' }
];
// Initialize weights with random values
this.weights = [
this.generateRandomWeights(3, 4), // Input to Hidden
this.generateRandomWeights(4, 2) // Hidden to Output
];
// Initialize biases
this.biases = [
Array(4).fill(0).map(() => Math.random() * 0.2 - 0.1), // Hidden layer biases
Array(2).fill(0).map(() => Math.random() * 0.2 - 0.1) // Output layer biases
];
// For animation purposes
this.activations = [
Array(3).fill(0), // Input activations
Array(4).fill(0), // Hidden layer activations
Array(2).fill(0) // Output activations
];
this.gradients = [
Array(3 * 4).fill(0), // Input to Hidden gradients
Array(4 * 2).fill(0) // Hidden to Output gradients
];
// Expected output for the sample
this.expectedOutput = [1, 0];
// Sample input
this.sampleInput = [0.8, 0.2, 0.5];
// Error
this.error = 0;
}
generateRandomWeights(inputSize, outputSize) {
const weights = [];
for (let i = 0; i < inputSize * outputSize; i++) {
weights.push(Math.random() * 0.4 - 0.2); // Random weights between -0.2 and 0.2
}
return weights;
}
// Activation functions
relu(x) {
return Math.max(0, x);
}
sigmoid(x) {
return 1 / (1 + Math.exp(-x));
}
// Forward pass
forwardPass() {
// Set input layer activations to sample input
this.activations[0] = [...this.sampleInput];
// Calculate hidden layer activations
for (let i = 0; i < this.layers[1].neurons; i++) {
let sum = this.biases[0][i];
for (let j = 0; j < this.layers[0].neurons; j++) {
const weightIdx = j * this.layers[1].neurons + i;
sum += this.activations[0][j] * this.weights[0][weightIdx];
}
this.activations[1][i] = this.relu(sum);
}
// Calculate output layer activations
for (let i = 0; i < this.layers[2].neurons; i++) {
let sum = this.biases[1][i];
for (let j = 0; j < this.layers[1].neurons; j++) {
const weightIdx = j * this.layers[2].neurons + i;
sum += this.activations[1][j] * this.weights[1][weightIdx];
}
this.activations[2][i] = this.sigmoid(sum);
}
// Calculate error (mean squared error)
this.error = 0;
for (let i = 0; i < this.layers[2].neurons; i++) {
const diff = this.activations[2][i] - this.expectedOutput[i];
this.error += diff * diff;
}
this.error /= this.layers[2].neurons;
return this.activations[2]; // Return output
}
// Calculate gradients (backward pass)
calculateGradients() {
// Output layer gradients
const outputDeltas = [];
for (let i = 0; i < this.layers[2].neurons; i++) {
const output = this.activations[2][i];
const target = this.expectedOutput[i];
// Derivative of loss with respect to output * derivative of sigmoid
outputDeltas.push((output - target) * output * (1 - output));
}
// Hidden to Output gradients
for (let i = 0; i < this.layers[1].neurons; i++) {
for (let j = 0; j < this.layers[2].neurons; j++) {
const weightIdx = i * this.layers[2].neurons + j;
this.gradients[1][weightIdx] = this.activations[1][i] * outputDeltas[j];
}
}
// Hidden layer deltas
const hiddenDeltas = Array(this.layers[1].neurons).fill(0);
for (let i = 0; i < this.layers[1].neurons; i++) {
let sum = 0;
for (let j = 0; j < this.layers[2].neurons; j++) {
const weightIdx = i * this.layers[2].neurons + j;
sum += this.weights[1][weightIdx] * outputDeltas[j];
}
// ReLU derivative is 1 if x > 0, otherwise 0
hiddenDeltas[i] = sum * (this.activations[1][i] > 0 ? 1 : 0);
}
// Input to Hidden gradients
for (let i = 0; i < this.layers[0].neurons; i++) {
for (let j = 0; j < this.layers[1].neurons; j++) {
const weightIdx = i * this.layers[1].neurons + j;
this.gradients[0][weightIdx] = this.activations[0][i] * hiddenDeltas[j];
}
}
return this.gradients;
}
// Update weights based on gradients
updateWeights(learningRate = 0.1) {
// Update weights using calculated gradients
for (let layerIdx = 0; layerIdx < this.weights.length; layerIdx++) {
for (let i = 0; i < this.weights[layerIdx].length; i++) {
this.weights[layerIdx][i] -= learningRate * this.gradients[layerIdx][i];
}
}
// Update biases (not shown in animation for simplicity)
// In a real implementation, we would update biases too
}
}
// Canvas resize functionality
function resizeCanvas() {
const container = canvas.parentElement;
canvas.width = container.clientWidth;
canvas.height = container.clientHeight;
// Redraw if already animating
if (animationState.network) {
drawNetwork(animationState.network);
}
}
// Initialize animation
function initAnimation() {
if (!canvas) return;
resizeCanvas();
window.addEventListener('resize', resizeCanvas);
// Create neural network
animationState.network = new NeuralNetwork();
// Draw initial state
drawNetwork(animationState.network);
// Update variables display
updateVariablesDisplay(animationState.network);
// Set button states
startButton.disabled = false;
pauseButton.disabled = true;
resetButton.disabled = true;
}
// Draw the neural network
function drawNetwork(network) {
if (!ctx) return;
// Clear canvas
ctx.clearRect(0, 0, canvas.width, canvas.height);
const padding = 50;
const width = canvas.width - padding * 2;
const height = canvas.height - padding * 2;
// Calculate neuron positions
const layers = network.layers;
const layerPositions = [];
for (let i = 0; i < layers.length; i++) {
const layerNeurons = [];
const x = padding + (width / (layers.length - 1)) * i;
for (let j = 0; j < layers[i].neurons; j++) {
const y = padding + (height / (layers[i].neurons + 1)) * (j + 1);
layerNeurons.push({ x, y });
}
layerPositions.push(layerNeurons);
}
// Draw connections
for (let layerIdx = 0; layerIdx < layers.length - 1; layerIdx++) {
for (let i = 0; i < layers[layerIdx].neurons; i++) {
for (let j = 0; j < layers[layerIdx + 1].neurons; j++) {
const weightIdx = i * layers[layerIdx + 1].neurons + j;
const weight = network.weights[layerIdx][weightIdx];
// Map weight to opacity for visualization
const normalizedWeight = Math.min(Math.abs(weight) * 5, 1);
// Set connection style based on the animation step
let connectionColor = '#ccc';
if (animationState.currentStep === 1) {
// Forward pass: blue
connectionColor = `rgba(52, 152, 219, ${normalizedWeight})`;
} else if (animationState.currentStep === 2) {
// Error calculation: red
if (layerIdx === network.weights.length - 1) {
connectionColor = `rgba(231, 76, 60, ${normalizedWeight})`;
} else {
connectionColor = `rgba(52, 152, 219, ${normalizedWeight})`;
}
} else if (animationState.currentStep === 3) {
// Backward pass: purple
connectionColor = `rgba(155, 89, 182, ${normalizedWeight})`;
} else if (animationState.currentStep === 4) {
// Weight update: green
const gradientNormalized = Math.min(Math.abs(network.gradients[layerIdx][weightIdx]) * 20, 1);
connectionColor = `rgba(46, 204, 113, ${gradientNormalized})`;
} else {
// Default state: gray with weight intensity
connectionColor = `rgba(150, 150, 150, ${normalizedWeight})`;
}
// Draw the connection
ctx.beginPath();
ctx.moveTo(layerPositions[layerIdx][i].x, layerPositions[layerIdx][i].y);
ctx.lineTo(layerPositions[layerIdx + 1][j].x, layerPositions[layerIdx + 1][j].y);
ctx.strokeStyle = connectionColor;
ctx.lineWidth = 2;
ctx.stroke();
}
}
}
// Draw neurons
for (let layerIdx = 0; layerIdx < layers.length; layerIdx++) {
for (let i = 0; i < layers[layerIdx].neurons; i++) {
const { x, y } = layerPositions[layerIdx][i];
// Set neuron style based on activation value
const activation = network.activations[layerIdx][i];
const activationColor = `rgba(52, 152, 219, ${Math.min(Math.max(activation, 0.2), 0.9)})`;
// Draw neuron
ctx.beginPath();
ctx.arc(x, y, 20, 0, Math.PI * 2);
ctx.fillStyle = activationColor;
ctx.fill();
ctx.strokeStyle = '#2980b9';
ctx.lineWidth = 2;
ctx.stroke();
// Draw neuron value
ctx.fillStyle = '#fff';
ctx.font = '12px Arial';
ctx.textAlign = 'center';
ctx.textBaseline = 'middle';
ctx.fillText(activation.toFixed(2), x, y);
// Draw layer labels
if (i === 0) {
ctx.fillStyle = '#333';
ctx.font = '14px Arial';
ctx.textAlign = 'center';
ctx.fillText(layers[layerIdx].type.toUpperCase(), x, y - 40);
}
}
}
}
// Update the variables display
function updateVariablesDisplay(network) {
const variablesContainer = document.getElementById('variables-container');
if (!variablesContainer) return;
let html = '';
// Different display based on animation step
switch (animationState.currentStep) {
case 1: // Forward Pass
html += `<div class="variable">Input: [${network.activations[0].map(v => v.toFixed(2)).join(', ')}]</div>`;
html += `<div class="variable">Hidden: [${network.activations[1].map(v => v.toFixed(2)).join(', ')}]</div>`;
html += `<div class="variable">Output: [${network.activations[2].map(v => v.toFixed(2)).join(', ')}]</div>`;
break;
case 2: // Error Calculation
html += `<div class="variable">Prediction: [${network.activations[2].map(v => v.toFixed(2)).join(', ')}]</div>`;
html += `<div class="variable">Target: [${network.expectedOutput.join(', ')}]</div>`;
html += `<div class="variable">Error: ${network.error.toFixed(4)}</div>`;
break;
case 3: // Backward Pass
html += `<div class="variable">Output Deltas:</div>`;
for (let i = 0; i < network.layers[2].neurons; i++) {
const output = network.activations[2][i];
const target = network.expectedOutput[i];
const delta = (output - target) * output * (1 - output);
html += `<div class="variable"> δ${i}: ${delta.toFixed(4)}</div>`;
}
break;
case 4: // Weight Updates
html += `<div class="variable">Selected Gradients:</div>`;
// Show just a few example gradients to avoid clutter
for (let layerIdx = 0; layerIdx < network.gradients.length; layerIdx++) {
const layerName = layerIdx === 0 ? 'Input→Hidden' : 'Hidden→Output';
html += `<div class="variable">${layerName}:</div>`;
// Show first few gradients as examples
for (let i = 0; i < Math.min(3, network.gradients[layerIdx].length); i++) {
html += `<div class="variable"> ∇w${i}: ${network.gradients[layerIdx][i].toFixed(4)}</div>`;
}
}
break;
default:
html += `<div class="variable">Click "Start Animation" to begin</div>`;
}
variablesContainer.innerHTML = html;
}
// Animation steps
const animationSteps = [
{
name: 'Starting',
description: 'Neural network in initial state. Click "Start Animation" to begin.'
},
{
name: 'Forward Pass',
description: 'Input data flows through the network to produce a prediction. Each neuron computes a weighted sum of its inputs, then applies an activation function.'
},
{
name: 'Error Calculation',
description: 'The network compares its prediction with the expected output to compute the error. This error measures how far off the prediction is.'
},
{
name: 'Backward Pass',
description: 'The error is propagated backward through the network, assigning responsibility to each weight for the prediction error.'
},
{
name: 'Weight Updates',
description: 'Weights are adjusted in proportion to their contribution to the error. Weights that contributed more to the error are adjusted more significantly.'
}
];
// Update step information display
function updateStepInfo(stepIndex) {
const stepName = document.getElementById('step-name');
const stepDescription = document.getElementById('step-description');
if (stepName && stepDescription && animationSteps[stepIndex]) {
stepName.textContent = animationSteps[stepIndex].name;
stepDescription.textContent = animationSteps[stepIndex].description;
}
}
// Animation loop
function animate(timestamp) {
if (!animationState.running) return;
// Calculate delta time for animation speed
const deltaTime = timestamp - animationState.lastTimestamp;
const interval = 3000 / animationState.speed; // Base interval divided by speed
if (deltaTime > interval || animationState.lastTimestamp === 0) {
animationState.lastTimestamp = timestamp;
// Progress through animation steps
if (animationState.currentStep === 0) {
// Initial state to forward pass
animationState.currentStep = 1;
animationState.network.forwardPass();
} else if (animationState.currentStep === 1) {
// Forward pass to error calculation
animationState.currentStep = 2;
} else if (animationState.currentStep === 2) {
// Error calculation to backward pass
animationState.currentStep = 3;
animationState.network.calculateGradients();
} else if (animationState.currentStep === 3) {
// Backward pass to weight updates
animationState.currentStep = 4;
} else if (animationState.currentStep === 4) {
// Weight updates to new forward pass
animationState.network.updateWeights(0.1);
animationState.currentStep = 1;
animationState.network.forwardPass();
}
// Update visuals
drawNetwork(animationState.network);
updateVariablesDisplay(animationState.network);
updateStepInfo(animationState.currentStep);
}
// Continue animation
animationState.animationFrameId = requestAnimationFrame(animate);
}
// Start animation
function startAnimation() {
if (!animationState.running) {
animationState.running = true;
animationState.lastTimestamp = 0;
animationState.animationFrameId = requestAnimationFrame(animate);
startButton.disabled = true;
pauseButton.disabled = false;
resetButton.disabled = false;
}
}
// Pause animation
function pauseAnimation() {
if (animationState.running) {
animationState.running = false;
if (animationState.animationFrameId) {
cancelAnimationFrame(animationState.animationFrameId);
}
startButton.disabled = false;
pauseButton.disabled = true;
resetButton.disabled = false;
}
}
// Reset animation
function resetAnimation() {
pauseAnimation();
animationState.currentStep = 0;
animationState.network = new NeuralNetwork();
drawNetwork(animationState.network);
updateVariablesDisplay(animationState.network);
updateStepInfo(animationState.currentStep);
startButton.disabled = false;
pauseButton.disabled = true;
resetButton.disabled = true;
}
// Control event listeners
startButton.addEventListener('click', startAnimation);
pauseButton.addEventListener('click', pauseAnimation);
resetButton.addEventListener('click', resetAnimation);
speedControl.addEventListener('input', () => {
animationState.speed = parseInt(speedControl.value, 10);
});
// Initialize the animation
initAnimation();
}); |