This repository has been archived by the owner on May 28, 2024. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 7
/
index.html
534 lines (490 loc) · 25.1 KB
/
index.html
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
<!doctype html>
<!-- Copyright 2016 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================-->
<html>
<head lang="en">
<link rel="shortcut icon" type="image/x-icon" href="favicon.ico" />
<meta charset="utf-8">
<meta name="viewport" content="width=1024">
<meta name="keywords" content="neural networks,machine learning,geoscience,geocomputing">
<meta property="og:type" content="article"/>
<meta property="og:title" content="Rocky Playground"/>
<meta property="og:description" content="Tinker with a neural network right here in your browser.">
<meta property="og:url" content="http://playground.geosci.ai"/>
<meta property="og:image" content="https://rocky-playground.s3.amazonaws.com/rocky-playground.png"/>
<meta name="twitter:card" value="summary_large_image">
<meta name="twitter:title" content="Rocky Playground">
<meta name="twitter:description" content="Tinker with a neural network right here in your browser.">
<meta name="twitter:url" content="http://playground.geosci.ai">
<meta name="twitter:image" content="https://rocky-playground.s3.amazonaws.com/rocky-playground.png">
<meta name="twitter:image:width" content="874">
<meta name="twitter:image:height" content="430">
<meta name="author" content="Daniel Smilkov and Shan Carter">
<title>A Rocky Neural Network Playground</title>
<link rel="stylesheet" href="bundle.css" type="text/css">
<link href="https://fonts.googleapis.com/css?family=Roboto:300,400,500|Material+Icons" rel="stylesheet" type="text/css">
<script src="lib.js"></script>
<!-- Should not need this -->
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/prism/1.5.0/themes/prism.min.css"/>
</head>
<body>
<!-- GitHub link -->
<a class="github-link" href="https://github.com/agilescientific/rocky-playground" title="Source on GitHub" target="_blank">
<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" viewBox="0 0 60.5 60.5" width="60" height="60">
<polygon class="bg" points="60.5,60.5 0,0 60.5,0 "/>
<path class="icon" d="M43.1,5.8c-6.6,0-12,5.4-12,12c0,5.3,3.4,9.8,8.2,11.4c0.6,0.1,0.8-0.3,0.8-0.6c0-0.3,0-1,0-2c-3.3,0.7-4-1.6-4-1.6c-0.5-1.4-1.3-1.8-1.3-1.8c-1.1-0.7,0.1-0.7,0.1-0.7c1.2,0.1,1.8,1.2,1.8,1.2c1.1,1.8,2.8,1.3,3.5,1c0.1-0.8,0.4-1.3,0.8-1.6c-2.7-0.3-5.5-1.3-5.5-5.9c0-1.3,0.5-2.4,1.2-3.2c-0.1-0.3-0.5-1.5,0.1-3.2c0,0,1-0.3,3.3,1.2c1-0.3,2-0.4,3-0.4c1,0,2,0.1,3,0.4c2.3-1.6,3.3-1.2,3.3-1.2c0.7,1.7,0.2,2.9,0.1,3.2c0.8,0.8,1.2,1.9,1.2,3.2c0,4.6-2.8,5.6-5.5,5.9c0.4,0.4,0.8,1.1,0.8,2.2c0,1.6,0,2.9,0,3.3c0,0.3,0.2,0.7,0.8,0.6c4.8-1.6,8.2-6.1,8.2-11.4C55.1,11.2,49.7,5.8,43.1,5.8z"/>
</svg>
</a>
<!-- Header -->
<header>
<h1 class="l--page">Tinker with a <b>neural network</b> <span class="optional">right here </span>in your browser.<br>Don’t worry, you can’t break it. Have fun!</h1>
</header>
<!-- Top Controls -->
<div id="top-controls">
<div class="container l--page">
<div class="timeline-controls">
<button class="mdl-button mdl-js-button mdl-button--icon ui-resetButton" id="reset-button" title="Reset the network">
<i class="material-icons">replay</i>
</button>
<button class="mdl-button mdl-js-button mdl-button--fab mdl-button--colored ui-playButton" id="play-pause-button" title="Run/Pause">
<i class="material-icons">play_arrow</i>
<i class="material-icons">pause</i>
</button>
<button class="mdl-button mdl-js-button mdl-button--icon ui-stepButton" id="next-step-button" title="Step">
<i class="material-icons">skip_next</i>
</button>
</div>
<div class="control">
<span class="label">Epoch</span>
<span class="value" id="iter-number"></span>
</div>
<div class="control ui-learningRate">
<label for="learningRate">Learning rate</label>
<div class="select">
<select id="learningRate">
<option value="0.00001">0.00001</option>
<option value="0.0001">0.0001</option>
<option value="0.001">0.001</option>
<option value="0.003">0.003</option>
<option value="0.01">0.01</option>
<option value="0.03">0.03</option>
<option value="0.1">0.1</option>
<option value="0.3">0.3</option>
<option value="1">1</option>
<option value="3">3</option>
<option value="10">10</option>
</select>
</div>
</div>
<div class="control ui-activation">
<label for="activations">Activation</label>
<div class="select">
<select id="activations">
<option value="linear">Linear</option>
<option value="sigmoid">Sigmoid</option>
<option value="tanh">Tanh</option>
<option value="relu">ReLU</option>
<option value="leakyrelu">Leaky ReLU</option>
<option value="elu">ELU</option>
<option value="swish">Swish</option>
<option value="gelu">GELU</option>
</select>
</div>
</div>
<div class="control ui-regularization">
<label for="regularizations">Regularization</label>
<div class="select">
<select id="regularizations">
<option value="none">None</option>
<option value="L1">L1</option>
<option value="L2">L2</option>
</select>
</div>
</div>
<div class="control ui-regularizationRate">
<label for="regularRate">Regularization rate</label>
<div class="select">
<select id="regularRate">
<option value="0">0</option>
<option value="0.001">0.001</option>
<option value="0.003">0.003</option>
<option value="0.01">0.01</option>
<option value="0.03">0.03</option>
<option value="0.1">0.1</option>
<option value="0.3">0.3</option>
<option value="1">1</option>
<option value="3">3</option>
<option value="10">10</option>
</select>
</div>
</div>
<div class="control ui-problem">
<label for="problem">Problem type</label>
<div class="select">
<select id="problem">
<option value="classification">Classification</option>
<option value="regression">Regression</option>
</select>
</div>
</div>
</div>
</div>
<!-- Main Part -->
<div id="main-part" class="l--page">
<!-- Data Column -->
<!-- Need even number of examples to keep spacing -->
<div class="column data">
<h4>
<span>Data</span>
</h4>
<!-- File uploader -->
<div class="ui-dataset">
<p>Upload your own JSON:</p>
<div class="dataset-list">
<div class="dataset-custom" title="Custom">
<canvas class="data-thumbnail" data-dataset="custom"></canvas>
</div>
<input type="file" id="files" name="files[]" />
<label class="mdl-button mdl-js-button" for="files" style="float:right">
<i class="material-icons">file_upload</i>
</label>
</div>
</div>
<div class="ui-dataset">
<p>Or use synthetic data:</p>
<div class="dataset-list">
<!-- Synthetic data -->
<div class="dataset" title="Linear">
<canvas class="data-thumbnail" data-dataset="linear"></canvas>
</div>
<div class="dataset" title="Diagonal">
<canvas class="data-thumbnail" data-dataset="diagonal"></canvas>
</div>
<div class="dataset" title="Exclusive or">
<canvas class="data-thumbnail" data-dataset="xor"></canvas>
</div>
<div class="dataset" title="Gaussian">
<canvas class="data-thumbnail" data-dataset="gauss"></canvas>
</div>
<div class="dataset" title="Circle">
<canvas class="data-thumbnail" data-dataset="circle"></canvas>
</div>
<div class="dataset" title="Spiral">
<canvas class="data-thumbnail" data-dataset="spiral"></canvas>
</div>
<div class="dataset" title="Moons">
<canvas class="data-thumbnail" data-dataset="moons"></canvas>
</div>
<div class="dataset" title="Plane">
<canvas class="data-thumbnail" data-regDataset="reg-plane"></canvas>
</div>
<div class="dataset" title="Multi gaussian">
<canvas class="data-thumbnail" data-regDataset="reg-gauss"></canvas>
</div>
</div>
<!-- Real data -->
<p>Or try real-world data:</p>
<div class="dataset-list">
<div class="dataset" title="Real data: sand vs shale">
<canvas class="data-thumbnail" data-dataset="rocks"></canvas>
</div>
<div class="dataset" title="Real data: poro-perm">
<canvas class="data-thumbnail" data-dataset="poroperm"></canvas>
</div>
<div class="dataset" title="Real data: poro-perm">
<canvas class="data-thumbnail" data-dataset="poroperm"></canvas>
</div>
<div class="dataset" title="Porosity">
<canvas class="data-thumbnail" data-regDataset="reg-porosity"></canvas>
</div>
<div class="dataset" title="DTS from DTP and RHOB">
<canvas class="data-thumbnail" data-regDataset="reg-dts"></canvas>
</div>
</div>
</div>
<div>
<div class="ui-percTrainData">
<label for="percTrainData">Train:test ratio: <span class="value">XX</span>%</label>
<p class="slider">
<input class="mdl-slider mdl-js-slider" type="range" id="percTrainData" min="10" max="90" step="10">
</p>
</div>
<div class="ui-noise">
<label for="noise">Noise: <span class="value">XX</span></label>
<p class="slider">
<input class="mdl-slider mdl-js-slider" type="range" id="noise" min="0" max="50" step="5">
</p>
</div>
<div class="ui-batchSize">
<label for="batchSize">Batch size: <span class="value">XX</span></label>
<p class="slider">
<input class="mdl-slider mdl-js-slider" type="range" id="batchSize" min="1" max="30" step="1">
</p>
</div>
<button class="basic-button" id="data-regen-button" title="Regenerate data">
Regenerate
</button>
</div>
</div>
<!-- Features Column -->
<div class="column features">
<h4>Features</h4>
<p>Which properties do you want to feed in?</p>
<div id="network">
<svg id="svg" width="510" height="450">
<defs>
<marker id="markerArrow" markerWidth="7" markerHeight="13" refX="1" refY="6" orient="auto" markerUnits="userSpaceOnUse">
<path d="M2,11 L7,6 L2,2" />
</marker>
</defs>
</svg>
<!-- Hover card -->
<div id="hovercard">
<div style="font-size:10px">Click anywhere to edit.</div>
<div><span class="type">Weight/Bias</span> is <span class="value">0.2</span><span><input type="number"/></span>.</div>
</div>
<div class="callout thumbnail">
<svg viewBox="0 0 30 30">
<defs>
<marker id="arrow" markerWidth="5" markerHeight="5" refx="5" refy="2.5" orient="auto" markerUnits="userSpaceOnUse">
<path d="M0,0 L5,2.5 L0,5 z"/>
</marker>
</defs>
<path d="M12,30C5,20 2,15 12,0" marker-end="url(#arrow)">
</svg>
<div class="label">
This is the output from one <b>neuron</b>. Hover to see it larger.
</div>
</div>
<div class="callout weights">
<svg viewBox="0 0 30 30">
<defs>
<marker id="arrow" markerWidth="5" markerHeight="5" refx="5" refy="2.5" orient="auto" markerUnits="userSpaceOnUse">
<path d="M0,0 L5,2.5 L0,5 z"/>
</marker>
</defs>
<path d="M12,30C5,20 2,15 12,0" marker-end="url(#arrow)">
</svg>
<div class="label">
The outputs are mixed with varying <b>weights</b>, shown by the thickness of the lines.
</div>
</div>
</div>
</div>
<!-- Hidden Layers Column -->
<div class="column hidden-layers">
<h4>
<div class="ui-numHiddenLayers">
<button id="add-layers" class="mdl-button mdl-js-button mdl-button--icon">
<i class="material-icons">add</i>
</button>
<button id="remove-layers" class="mdl-button mdl-js-button mdl-button--icon">
<i class="material-icons">remove</i>
</button>
</div>
<span id="num-layers"></span>
<span id="layers-label"></span>
</h4>
<div class="bracket"></div>
</div>
<!-- Output Column -->
<div class="column output">
<h4>Output</h4>
<div class="metrics">
<div class="output-stats ui-percTrainData">
<span>Test loss</span>
<div class="value" id="loss-test"></div>
</div>
<div class="output-stats train">
<span>Training loss</span>
<div class="value" id="loss-train"></div>
</div>
<div id="linechart"></div>
</div>
<div id="heatmap"></div>
<div style="float:left;margin-top:20px">
<div style="display:flex; align-items:center;">
<!-- Gradient color scale -->
<div class="label" style="width:105px; margin-right: 10px">
Colors shows data, neuron and weight values.
</div>
<svg width="150" height="30" id="colormap">
<defs>
<linearGradient id="gradient" x1="0%" y1="100%" x2="100%" y2="100%">
<stop offset="0%" stop-color="#325396" stop-opacity="1"></stop>
<stop offset="50%" stop-color="#e8eaeb" stop-opacity="1"></stop>
<stop offset="100%" stop-color="#16afca" stop-opacity="1"></stop>
</linearGradient>
</defs>
<g class="core" transform="translate(3, 0)">
<rect width="144" height="10" style="fill: url('#gradient');"></rect>
</g>
</svg>
</div>
<br/>
<div style="display:flex;">
<label class="ui-showTestData mdl-checkbox mdl-js-checkbox mdl-js-ripple-effect" for="show-test-data">
<input type="checkbox" id="show-test-data" class="mdl-checkbox__input" checked>
<span class="mdl-checkbox__label label">Show test data</span>
</label>
<label class="ui-discretize mdl-checkbox mdl-js-checkbox mdl-js-ripple-effect" for="discretize">
<input type="checkbox" id="discretize" class="mdl-checkbox__input" checked>
<span class="mdl-checkbox__label label">Discretize output</span>
</label>
</div>
</div>
</div>
</div>
<!-- Authored by https://github.com/jameshfisher and adapted by Matt -->
<div class="l--page">
<div class="column">
<div class="card">
<h2>This neural network as a Python function</h2>
<pre><code id="network-as-python" class="language-python"></code></pre>
</div>
</div>
</div>
<!-- More -->
<div class="more">
<button class="mdl-button mdl-js-button mdl-button--fab">
<i class="material-icons">keyboard_arrow_down</i>
</button>
</div>
<!-- Article -->
<article id="article-text">
<div class="l--body">
<h2>What is a neural network?</h2>
<p>It’s a technique for building a computer program that learns from data, based very loosely on how brains work. Software “neurons” are connected together, allowing them to send messages to each other. The network is asked to solve a problem, which it attempts to do over and over, each time strengthening the connections that lead to success and diminishing those that lead to failure. For a more detailed introduction to neural networks, Michael Nielsen’s <a href="http://neuralnetworksanddeeplearning.com/index.html">Neural Networks and Deep Learning</a> is a good place to start.</p>
</div>
<div class="l--body">
<h2>More about the datasets</h2>
<p>The data points are scaled and dimensionless, so it might be hard to tell what they represent. So here's a brief description.</p>
<h3>Classification datasets</h3>
<div class="tutorial">
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-dataset-tuto="linear"></canvas>
</div>
<p class="tutorial-text"><strong>Linear</strong> – A data set that is linearly seperable along the X<sub>1</sub> dimension. You'll only need one feature and one neuron to learn this relationship.</sub></p>
</div>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-dataset-tuto="diagonal"></canvas>
</div>
<p class="tutorial-text"><strong>Diagonal</strong> – The two classes are clustered along a diagonal, so you'll need more than one feature to build an adequate classifier. Note that the training data does not span the domain of the data, whereas the validation data does — so the variance is always high on this dataset.</p>
</div>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-dataset-tuto="xor"></canvas>
</div>
<p class="tutorial-text"><strong>Exclusive Or (XOR)</strong> – This is a classic problem in neural network research. It is the simplest non-linearly separable classification task that exists.</p>
</div>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-dataset-tuto="gauss"></canvas>
</div>
<p class="tutorial-text"><strong>Gaussian</strong> – A classification problem with clusters (or blobs) each represented by a Gaussian distribution. Change the noise level to broaden or narrow the distribution.</p>
</div>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-dataset-tuto="circle"></canvas>
</div>
<p class="tutorial-text"><strong>Circle</strong> – The instances of one class encircle the instances of the other.</p>
</div>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-dataset-tuto="spiral"></canvas>
</div>
<p class="tutorial-text"><strong>Spiral</strong> – Two classes spiraling around each other.</p>
</div>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-dataset-tuto="moons"></canvas>
</div>
<p class="tutorial-text"><strong>Moons</strong> – Two interleaving moons, or croissants if you prefer. Inspired by the <code>sklearn</code> dataset.</code></p>
</div>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-dataset-tuto="rocks"></canvas>
</div>
<p class="tutorial-text"><strong>Real data: sand vs shale</strong> – This is P-wave velocity (X<sub>1</sub>) and bulk density (X<sub>2</sub>) values for a selection of sandstones (cyan) and shales (dark blue) from the <a href="https://subsurfwiki.org/wiki/Rock_Property_Catalog">Rock Property Catalog</a>. A standard scalar has been applied to the features, then multipled by 2.</a></p>
</div>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-dataset-tuto="poroperm"></canvas>
</div>
<p class="tutorial-text"><strong>Real data: poro-perm</strong> – Porosity (X<sub>1</sub>) and the base-10 logarithm of permeability (X<sub>2</sub>) of Oligocene sandstones. Medium-to-coarse grained sands are in cyan, fine-to-medium grained sands are in dark blue. Dataset is modified from Taylor et al. 1993, dataset number 64 in the <a href="https://pubs.usgs.gov/of/2003/ofr-03-420/ofr-03-420.html">USGS report 03-420</a> of Porosity and Permeability from Core Plugs in Siliclastic Rocks.</p>
</div>
<h3>Regression datasets</h3>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-regDataset-tuto="reg-plane"></canvas>
</div>
<p class="tutorial-text"><strong>Plane</strong> – Data coordinates sampling a dipping planar surface.</p>
</div>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-regDataset-tuto="reg-gauss"></canvas>
</div>
<p class="tutorial-text"><strong>Multi-Gaussian</strong> – multiple clusters with Gaussian spatial distributions.</p>
</div>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-regDataset-tuto="reg-porosity"></canvas>
</div>
<p class="tutorial-text"><strong>Real data: Porosity</strong> – This is a small dataset representing a porosity map. It is from Geoff Bohling at the Kansas Geological Survey, but we can no longer find the data online.</p>
</div>
<div class="tutorial-item">
<div class="dataset">
<canvas class="data-thumbnail" data-regDataset-tuto="reg-dts"></canvas>
</div>
<p class="tutorial-text"><strong>Real data: DTS from DTP and RHOB</strong> – Predicting S-wave sonic wireline measurements from compressional velocity (DTP) (X<sub>1</sub>) and bulk density (X<sub>2</sub>). Data has been downsampled from well R-39 offshore Nova Scotia, available from the <a href="https://www.cnsopb.ns.ca/">CNSOPB</a>.</p>
</div>
</div>
</div> <!-- end of tutorial -->
<div class="l--body">
<h2>How do I make my own datasets?</h2>
<p>You will need to define a machine learning task with 2 features and one target. About 400 records is ideal — split 50% positive and 50% negative classes for a classification problem. The features will be scaled using Z-score standardization (multiplied by 2 for complicated reasons). The target should be -1 and 1 for classification, or in the range [-1, 1] for regression. <a href="https://github.com/agilescientific/rocky-playground/blob/master/notebooks/Data_for_Playground.ipynb">Here's an example of how to prepare data.</a></p>
<p>Once you have a JSON file, you can upload it using the button with the <i class="material-icons">file_upload</i> icon (top left).</p>
</div>
<div class="l--body">
<h2>Credits</h2>
<p>This fork of <a href="https://playground.tensorflow.org/">the TensorFlow Playground</a> was adapted by Matt Hall and Evan Bianco at <a href="https://agilescientific.com/">Agile Scientific</a>. It's one of Agile's <a href="https://geosci.ai">geosci.ai</a> experiments.</p>
<p>Here's some of what we've added or changed:</p>
<ul>
<li>Added several new real and synthetic datasets, with descriptions.</li>
<li>More activation functions to try, including ELU and Swish.</li>
<li>Change the regularization during training and watch the weights.</li>
<li>Upload your own datasets!</li>
<li>See the expression of the network in Python code.</li>
<li>Lots of small bug fixes and cosmetic changes.</li>
</ul>
<p>
The original app was was created by Daniel Smilkov and Shan Carter. It was a continuation of many people’s previous work — most notably Andrej Karpathy’s <a href="http://cs.stanford.edu/people/karpathy/convnetjs/demo/classify2d.html">convnet.js demo</a>
and Chris Olah’s <a href="http://colah.github.io/posts/2014-03-NN-Manifolds-Topology/">articles</a> about neural networks.
Many thanks also to D. Sculley for help with the original idea and to Fernanda Viégas and Martin Wattenberg and the rest of the
<a href="https://research.google.com/bigpicture/">Big Picture</a> and <a href="https://research.google.com/teams/brain/">Google Brain</a> teams for feedback and guidance.
</p>
<p>Several of our changes were inspired or pioneered by <a href="https://dcato98.github.io/playground">David Cato's</a> fork of the project, especially the Swish function, the code expression, and the on-the-fly regularization tweaking. Kudos and thanks to him and his collaborators!</p>
</div>
</article>
<!-- Footer -->
<footer>
<div class="l--body">
<a href="https://www.agilescientific.com/" class="logo">
<img width=120 src="https://rocky-playground.s3.amazonaws.com/Agile_logo_wo_web_rect.png" alt="Agile Scientific">
</a>
</div>
</footer>
<script src="bundle.js"></script>
</body>
</html>