Skip to content

Commit e0ff8e7

Browse files
xieguigangxieguigang
committed
generates the html help document
1 parent 129ac22 commit e0ff8e7

File tree

787 files changed

+60041
-2
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

787 files changed

+60041
-2
lines changed

‎.github/build-githubpage.R

+5-2
Original file line numberDiff line numberDiff line change
@@ -15,7 +15,10 @@ for(dll_name in clr_modules) {
1515

1616
# generates the html help documents
1717
for(pkg_namespace in package_utils::parseDll(dll = `${app_dir}/${dll_name}`)) {
18-
Rdocuments(pkg_namespace, outputdir = `${vignettes}/${basename(dll_name)}`,
19-
package = "R");
18+
try({
19+
Rdocuments(pkg_namespace,
20+
outputdir = `${vignettes}/${basename(dll_name)}/${[pkg_namespace]::namespace}`,
21+
package = "R");
22+
});
2023
}
2124
}

‎.github/build-githubpage.cmd

+2
Original file line numberDiff line numberDiff line change
@@ -4,3 +4,5 @@ SET R_HOME=/GCModeller/src/R-sharp/App/net6.0-windows
44
SET R_ENV="%R_HOME%/R#.exe"
55

66
%R_ENV% ./build-githubpage.R
7+
8+
pause

‎vignettes/MLkit/CNN.html

+313
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,313 @@
1+
<!DOCTYPE html><html lang="zh-CN">
2+
<head>
3+
<meta http-equiv="content-type" content="text/html; charset=UTF-8" />
4+
<meta http-equiv="X-UA-Compatible" content="IE=Edge" />
5+
<meta charset="utf-8" />
6+
<meta name="viewport" content="width=device-width, minimum-scale=1.0, maximum-scale=1.0" />
7+
<title>CNN</title>
8+
<meta name="author" content="[email protected]" />
9+
<meta name="copyright" content="SMRUCC genomics Copyright (c) 2022" />
10+
<meta name="keywords" content="R#; CNN; MLkit" />
11+
<meta name="generator" content="https://github.com/rsharp-lang" />
12+
<meta name="theme-color" content="#333" />
13+
<meta name="description" content="feed-forward phase of deep Convolutional Neural Networks..." />
14+
<meta class="foundation-data-attribute-namespace" />
15+
<meta class="foundation-mq-xxlarge" />
16+
<meta class="foundation-mq-xlarge" />
17+
<meta class="foundation-mq-large" />
18+
<meta class="foundation-mq-medium" />
19+
<meta class="foundation-mq-small" />
20+
<meta class="foundation-mq-topbar" />
21+
<style>
22+
23+
.table-three-line {
24+
border-collapse:collapse; /* 关键属性：合并表格内外边框(其实表格边框有2px，外面1px，里面还有1px哦) */
25+
border:solid #000000; /* 设置边框属性；样式(solid=实线)、颜色(#999=灰) */
26+
border-width:2px 0 2px 0px; /* 设置边框状粗细：上 右 下 左 = 对应：1px 0 0 1px */
27+
}
28+
.left-1{
29+
border:solid #000000;border-width:1px 1px 2px 0px;padding:2px;
30+
font-weight:bolder;
31+
}
32+
.right-1{
33+
border:solid #000000;border-width:1px 0px 2px 1px;padding:2px;
34+
font-weight:bolder;
35+
}
36+
.mid-1{
37+
border:solid #000000;border-width:1px 1px 2px 1px;padding:2px;
38+
font-weight:bolder;
39+
}
40+
.left{
41+
border:solid #000000;border-width:1px 1px 1px 0px;padding:2px;
42+
}
43+
.right{
44+
border:solid #000000;border-width:1px 0px 1px 1px;padding:2px;
45+
}
46+
.mid{
47+
border:solid #000000;border-width:1px 1px 1px 1px;padding:2px;
48+
}
49+
table caption {font-size:14px;font-weight:bolder;}
50+
</style>
51+
</head>
52+
<body>
53+
<table width="100%" summary="page for {CNN}">
54+
<tbody>
55+
<tr>
56+
<td>{CNN}</td>
57+
<td style="text-align: right;">R# Documentation</td>
58+
</tr>
59+
</tbody>
60+
</table>
61+
<h1>CNN</h1>
62+
<hr />
63+
<p style=" font-size: 1.125em; line-height: .8em; margin-left: 0.5%; background-color: #fbfbfb; padding: 24px; ">
64+
<code>
65+
<span style="color: blue;">require</span>(<span style="color: black; font-weight: bold;">R</span>);
66+
<br /><br /><span style="color: green;">#' feed-forward phase of deep Convolutional Neural Networks</span><br /><span style="color: blue;">imports</span><span style="color: brown"> "CNN"</span><span style="color: blue;"> from</span><span style="color: brown"> "MLkit"</span>;
67+
</code>
68+
</p>
69+
<p><p>feed-forward phase of deep Convolutional Neural Networks</p></p>
70+
<blockquote>
71+
<p style="font-style: italic; font-size: 0.9em;">
72+
<p>feed-forward phase of deep Convolutional Neural Networks</p>
73+
</p>
74+
</blockquote>
75+
<div id="main-wrapper">
76+
<table class="table-three-line">
77+
<tbody><tr>
78+
<td id="n_threads">
79+
<a href="./CNN/n_threads.html">n_threads</a>
80+
</td>
81+
<td><p>get/set of the CNN parallel thread number</p></td>
82+
</tr>
83+
<tr>
84+
<td id="cnn">
85+
<a href="./CNN/cnn.html">cnn</a>
86+
</td>
87+
<td><p>Create a new CNN model</p>
88+
89+
<p>Convolutional neural network (CNN) is a regularized type of feed-forward<br />
90+
neural network that learns feature engineering by itself via filters <br />
91+
(or kernel) optimization. Vanishing gradients and exploding gradients, <br />
92+
seen during backpropagation in earlier neural networks, are prevented by <br />
93+
using regularized weights over fewer connections.</p></td>
94+
</tr>
95+
<tr>
96+
<td id="input_layer">
97+
<a href="./CNN/input_layer.html">input_layer</a>
98+
</td>
99+
<td><p>The input layer is a simple layer that will pass the data though and<br />
100+
create a window into the full training data set. So for instance if<br />
101+
we have an image of size 28x28x1 which means that we have 28 pixels<br />
102+
in the x axle and 28 pixels in the y axle and one color (gray scale),<br />
103+
then this layer might give you a window of another size example 24x24x1<br />
104+
that is randomly chosen in order to create some distortion into the<br />
105+
dataset so the algorithm don't over-fit the training.</p></td>
106+
</tr>
107+
<tr>
108+
<td id="regression_layer">
109+
<a href="./CNN/regression_layer.html">regression_layer</a>
110+
</td>
111+
<td></td>
112+
</tr>
113+
<tr>
114+
<td id="conv_layer">
115+
<a href="./CNN/conv_layer.html">conv_layer</a>
116+
</td>
117+
<td><p>This layer uses different filters to find attributes of the data that<br />
118+
affects the result. As an example there could be a filter to find<br />
119+
horizontal edges in an image.</p></td>
120+
</tr>
121+
<tr>
122+
<td id="conv_transpose_layer">
123+
<a href="./CNN/conv_transpose_layer.html">conv_transpose_layer</a>
124+
</td>
125+
<td></td>
126+
</tr>
127+
<tr>
128+
<td id="lrn_layer">
129+
<a href="./CNN/lrn_layer.html">lrn_layer</a>
130+
</td>
131+
<td><p>This layer is useful when we are dealing with ReLU neurons. Why is that?<br />
132+
Because ReLU neurons have unbounded activations and we need LRN to normalize<br />
133+
that. We want to detect high frequency features with a large response. If we<br />
134+
normalize around the local neighborhood of the excited neuron, it becomes even<br />
135+
more sensitive as compared to its neighbors.</p>
136+
137+
<p>At the same time, it will dampen the responses that are uniformly large in any<br />
138+
given local neighborhood. If all the values are large, then normalizing those<br />
139+
values will diminish all of them. So basically we want to encourage some kind<br />
140+
of inhibition and boost the neurons with relatively larger activations. This<br />
141+
has been discussed nicely in Section 3.3 of the original paper by Krizhevsky et al.</p></td>
142+
</tr>
143+
<tr>
144+
<td id="tanh_layer">
145+
<a href="./CNN/tanh_layer.html">tanh_layer</a>
146+
</td>
147+
<td><p>Implements Tanh nonlinearity elementwise x to tanh(x)<br />
148+
so the output is between -1 and 1.</p></td>
149+
</tr>
150+
<tr>
151+
<td id="softmax_layer">
152+
<a href="./CNN/softmax_layer.html">softmax_layer</a>
153+
</td>
154+
<td><p>[*loss_layers] This layer will squash the result of the activations in the fully<br />
155+
connected layer and give you a value of 0 to 1 for all output activations.</p></td>
156+
</tr>
157+
<tr>
158+
<td id="relu_layer">
159+
<a href="./CNN/relu_layer.html">relu_layer</a>
160+
</td>
161+
<td><p>This is a layer of neurons that applies the non-saturating activation<br />
162+
function f(x)=max(0,x). It increases the nonlinear properties of the<br />
163+
decision function and of the overall network without affecting the<br />
164+
receptive fields of the convolution layer.</p></td>
165+
</tr>
166+
<tr>
167+
<td id="leaky_relu_layer">
168+
<a href="./CNN/leaky_relu_layer.html">leaky_relu_layer</a>
169+
</td>
170+
<td></td>
171+
</tr>
172+
<tr>
173+
<td id="maxout_layer">
174+
<a href="./CNN/maxout_layer.html">maxout_layer</a>
175+
</td>
176+
<td><p>Implements Maxout nonlinearity that computes x to max(x)<br />
177+
where x is a vector of size group_size. Ideally of course,<br />
178+
the input size should be exactly divisible by group_size</p></td>
179+
</tr>
180+
<tr>
181+
<td id="sigmoid_layer">
182+
<a href="./CNN/sigmoid_layer.html">sigmoid_layer</a>
183+
</td>
184+
<td><p>Implements Sigmoid nonlinearity elementwise x to 1/(1+e^(-x))<br />
185+
so the output is between 0 and 1.</p></td>
186+
</tr>
187+
<tr>
188+
<td id="pool_layer">
189+
<a href="./CNN/pool_layer.html">pool_layer</a>
190+
</td>
191+
<td><p>This layer will reduce the dataset by creating a smaller zoomed out<br />
192+
version. In essence you take a cluster of pixels take the sum of them<br />
193+
and put the result in the reduced position of the new image.</p></td>
194+
</tr>
195+
<tr>
196+
<td id="dropout_layer">
197+
<a href="./CNN/dropout_layer.html">dropout_layer</a>
198+
</td>
199+
<td><p>This layer will remove some random activations in order to<br />
200+
defeat over-fitting.</p></td>
201+
</tr>
202+
<tr>
203+
<td id="full_connected_layer">
204+
<a href="./CNN/full_connected_layer.html">full_connected_layer</a>
205+
</td>
206+
<td><p>Neurons in a fully connected layer have full connections to all<br />
207+
activations in the previous layer, as seen in regular Neural Networks.<br />
208+
Their activations can hence be computed with a matrix multiplication<br />
209+
followed by a bias offset.</p></td>
210+
</tr>
211+
<tr>
212+
<td id="gaussian_layer">
213+
<a href="./CNN/gaussian_layer.html">gaussian_layer</a>
214+
</td>
215+
<td></td>
216+
</tr>
217+
<tr>
218+
<td id="sample_dataset">
219+
<a href="./CNN/sample_dataset.html">sample_dataset</a>
220+
</td>
221+
<td></td>
222+
</tr>
223+
<tr>
224+
<td id="sample_dataset.image">
225+
<a href="./CNN/sample_dataset.image.html">sample_dataset.image</a>
226+
</td>
227+
<td></td>
228+
</tr>
229+
<tr>
230+
<td id="auto_encoder">
231+
<a href="./CNN/auto_encoder.html">auto_encoder</a>
232+
</td>
233+
<td></td>
234+
</tr>
235+
<tr>
236+
<td id="training">
237+
<a href="./CNN/training.html">training</a>
238+
</td>
239+
<td><p>Do CNN network model training</p></td>
240+
</tr>
241+
<tr>
242+
<td id="ada_delta">
243+
<a href="./CNN/ada_delta.html">ada_delta</a>
244+
</td>
245+
<td><p>Adaptive delta will look at the differences between the expected result and the current result to train the network.</p></td>
246+
</tr>
247+
<tr>
248+
<td id="ada_grad">
249+
<a href="./CNN/ada_grad.html">ada_grad</a>
250+
</td>
251+
<td><p>The adaptive gradient trainer will over time sum up the square of<br />
252+
the gradient and use it to change the weights.</p></td>
253+
</tr>
254+
<tr>
255+
<td id="adam">
256+
<a href="./CNN/adam.html">adam</a>
257+
</td>
258+
<td><p>Adaptive Moment Estimation is an update to RMSProp optimizer. In this running average of both the<br />
259+
gradients and their magnitudes are used.</p></td>
260+
</tr>
261+
<tr>
262+
<td id="nesterov">
263+
<a href="./CNN/nesterov.html">nesterov</a>
264+
</td>
265+
<td><p>Another extension of gradient descent is due to Yurii Nesterov from 1983,[7] and has been subsequently generalized</p></td>
266+
</tr>
267+
<tr>
268+
<td id="sgd">
269+
<a href="./CNN/sgd.html">sgd</a>
270+
</td>
271+
<td><p>Stochastic gradient descent (often shortened in SGD), also known as incremental gradient descent, is a<br />
272+
stochastic approximation of the gradient descent optimization method for minimizing an objective function<br />
273+
that is written as a sum of differentiable functions. In other words, SGD tries to find minimums or<br />
274+
maximums by iteration.</p></td>
275+
</tr>
276+
<tr>
277+
<td id="window_grad">
278+
<a href="./CNN/window_grad.html">window_grad</a>
279+
</td>
280+
<td><p>This is AdaGrad but with a moving window weighted average<br />
281+
so the gradient is not accumulated over the entire history of the run.<br />
282+
it's also referred to as Idea #1 in Zeiler paper on AdaDelta.</p></td>
283+
</tr>
284+
<tr>
285+
<td id="predict">
286+
<a href="./CNN/predict.html">predict</a>
287+
</td>
288+
<td></td>
289+
</tr>
290+
<tr>
291+
<td id="CeNiN">
292+
<a href="./CNN/CeNiN.html">CeNiN</a>
293+
</td>
294+
<td><p>load a CNN model from file</p></td>
295+
</tr>
296+
<tr>
297+
<td id="detectObject">
298+
<a href="./CNN/detectObject.html">detectObject</a>
299+
</td>
300+
<td><p>classify a object from a given image data</p></td>
301+
</tr>
302+
<tr>
303+
<td id="saveModel">
304+
<a href="./CNN/saveModel.html">saveModel</a>
305+
</td>
306+
<td><p>save the CNN model into a binary data file</p></td>
307+
</tr></tbody>
308+
</table>
309+
</div>
310+
<hr />
311+
<div style="text-align: center;">[<a href="../index.html">Document Index</a>]</div>
312+
</body>
313+
</html>

0 commit comments

Comments
 (0)