|
|
@@ -0,0 +1,101 @@
|
|
|
+<!DOCTYPE html>
|
|
|
+<html>
|
|
|
+
|
|
|
+<head>
|
|
|
+ <title>Brain JS - LSTM Math</title>
|
|
|
+ <meta charset="utf-8" />
|
|
|
+ <meta http-equiv="Content-Type"
|
|
|
+ content="text/html; charset=utf-8" />
|
|
|
+ <style>
|
|
|
+ #network {
|
|
|
+ background: #ccc;
|
|
|
+ border: 1px solid red;
|
|
|
+ width: 640px;
|
|
|
+ height: 480px;
|
|
|
+ }
|
|
|
+ </style>
|
|
|
+</head>
|
|
|
+
|
|
|
+<body>
|
|
|
+ <h1>Brain JS - LSTM Math</h1>
|
|
|
+ <p>
|
|
|
+ In this case, we test the math adding calcuation by LSTM model.
|
|
|
+ </p>
|
|
|
+ <h2>Test results</h2>
|
|
|
+ <div id="input">
|
|
|
+ <p>Input a math problem, like "1+2=" or "9+7=" in the below input box.</p>
|
|
|
+ <input type="text"
|
|
|
+ id="problem"></input>
|
|
|
+ <button id="calcuate">Calculate</button>
|
|
|
+ <button id="testall">Test all once</button>
|
|
|
+ </div>
|
|
|
+ <div id="output"></div>
|
|
|
+ <h2>Network</h2>
|
|
|
+ <div id="network"></div>
|
|
|
+</body>
|
|
|
+<script src="brain-browser.js"></script>
|
|
|
+<script type="text/javascript">
|
|
|
+const svg_config = {
|
|
|
+ height: 480,
|
|
|
+ width: 640,
|
|
|
+ radius: 10,
|
|
|
+};
|
|
|
+
|
|
|
+// create a simple recurrent neural network
|
|
|
+const net = new brain.recurrent.LSTM();
|
|
|
+
|
|
|
+// used to build list below
|
|
|
+const mathProblemsSet = new Set();
|
|
|
+for (let i = 0; i < 10; i++) {
|
|
|
+ for (let j = 0; j < 10; j++) {
|
|
|
+ mathProblemsSet.add(`${i}+${j}=${i + j}`);
|
|
|
+ mathProblemsSet.add(`${j}+${i}=${i + j}`);
|
|
|
+ }
|
|
|
+}
|
|
|
+const mathProblems = Array.from(mathProblemsSet);
|
|
|
+console.log(mathProblems);
|
|
|
+
|
|
|
+net.train(mathProblems, { log: true, errorThresh: 0.03 });
|
|
|
+
|
|
|
+
|
|
|
+let errors = 0;
|
|
|
+document.getElementById("calcuate").onclick = () => {
|
|
|
+ const input = document.getElementById("problem").value;
|
|
|
+ const output = net.run(input)
|
|
|
+
|
|
|
+ const predictedMathProblem = input + output;
|
|
|
+ const error = mathProblems.indexOf(predictedMathProblem) <= -1 ? 1 : 0;
|
|
|
+ errors += error;
|
|
|
+
|
|
|
+ document.getElementById("output").innerHTML =
|
|
|
+ `Output :${input}${output}` +
|
|
|
+ `<br/>${error?'Error':'Correct'}, total error ${errors}`;
|
|
|
+};
|
|
|
+
|
|
|
+// Batch test for all problems
|
|
|
+document.getElementById("testall").onclick = () => {
|
|
|
+ let results = 'Test all items:<br/><ul>';
|
|
|
+ for (let i = 0; i < mathProblems.length; i++) {
|
|
|
+ const input = mathProblems[i].split('=')[0] + '=';
|
|
|
+ const output = net.run(input);
|
|
|
+ const predictedMathProblem = input + output;
|
|
|
+ const error = mathProblems.indexOf(predictedMathProblem) <= -1 ? 1 : 0;
|
|
|
+ errors += error;
|
|
|
+ results += '<li>';
|
|
|
+ results += input + output;
|
|
|
+ results += error ? ' - error' : '';
|
|
|
+ results += '</li>';
|
|
|
+ }
|
|
|
+ results +='</ul>';
|
|
|
+
|
|
|
+ results = `Batch test with ${errors} errors <br/>`+ results;
|
|
|
+
|
|
|
+ document.getElementById("output").innerHTML = results;
|
|
|
+};
|
|
|
+document.getElementById("network").innerHTML = brain.utilities.toSVG(
|
|
|
+ net,
|
|
|
+ svg_config
|
|
|
+);
|
|
|
+</script>
|
|
|
+
|
|
|
+</html>
|
