wehub-resource-sync
48 lines
2.0 KiB
HTML
48 lines
2.0 KiB
HTML
<!doctype html>
|
|
<html lang="en">
|
|
<head>
|
|
<meta charset="utf-8">
|
|
<meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1">
|
|
<title>ConvNetJS Trainer Comparison on MNIST</title>
|
|
<meta name="description" content="">
|
|
<meta name="author" content="">
|
|
<link rel="stylesheet" href="css/style.css">
|
|
|
|
<script src="js/jquery-1.8.3.min.js"></script>
|
|
<script src="../build/vis.js"></script>
|
|
<script src="../build/util.js"></script>
|
|
<script src="../build/convnet.js"></script>
|
|
<script src="mnist/mnist_labels.js"></script>
|
|
<script src="js/trainers.js"></script>
|
|
|
|
</head>
|
|
<body>
|
|
<div id="wrap">
|
|
<h2 style="text-align: center;"><a href="http://cs.stanford.edu/people/karpathy/convnetjs/">ConvNetJS</a> Trainer demo on MNIST</h2>
|
|
<h1>Description</h1>
|
|
<p>
|
|
This demo lets you evaluate multiple trainers against each other on MNIST. By default I've set up a little benchmark that puts SGD/SGD with momentum/Adam/Adagrad/Adadelta/Nesterov against each other. For reference math and explanations on these refer to Matthew Zeiler's <a href="http://www.matthewzeiler.com/pubs/googleTR2012/googleTR2012.pdf">Adadelta paper</a> (Windowgrad is Idea #1 in the paper). In my own experience, Adagrad/Adadelta are "safer" because they don't depend so strongly on setting of learning rates (with Adadelta being slightly better), but well-tuned SGD+Momentum almost always converges faster and at better final values.
|
|
</p>
|
|
<p>Report questions/bugs/suggestions to <a href="https://twitter.com/karpathy">@karpathy</a>.</p>
|
|
|
|
<textarea id="layerdef" style="width:100%; height:400px;">
|
|
</textarea>
|
|
<br /><br />
|
|
<input id="buttontp" type="submit" value="re-run" onclick="reload();" style="width: 300px; height: 50px;"/>
|
|
|
|
<h1>Loss vs. Number of examples seen</h1>
|
|
<canvas id="lossgraph" width="800" height="400"></canvas>
|
|
|
|
<h1>Testing Accuracy vs. Number of examples seen</h1>
|
|
<canvas id="testaccgraph" width="800" height="400"></canvas>
|
|
|
|
<h1>Training Accuracy vs. Number of examples seen</h1>
|
|
<canvas id="trainaccgraph" width="800" height="400"></canvas>
|
|
|
|
</div>
|
|
</body>
|
|
</html>
|
|
|
|
|
|
|