Skip to content

Commit

Permalink
generates the html help document
Browse files Browse the repository at this point in the history
  • Loading branch information
@xieguigang
xieguigang committed Dec 28, 2023
1 parent 129ac22 commit e0ff8e7
Show file tree
Hide file tree
Showing 787 changed files with 60,041 additions and 2 deletions.
7 changes: 5 additions & 2 deletions .github/build-githubpage.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,10 @@ for(dll_name in clr_modules) {

# generates the html help documents
for(pkg_namespace in package_utils::parseDll(dll = `${app_dir}/${dll_name}`)) {
Rdocuments(pkg_namespace, outputdir = `${vignettes}/${basename(dll_name)}`,
package = "R");
try({
Rdocuments(pkg_namespace,
outputdir = `${vignettes}/${basename(dll_name)}/${[pkg_namespace]::namespace}`,
package = "R");
});
}
}
2 changes: 2 additions & 0 deletions .github/build-githubpage.cmd
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4,3 +4,5 @@ SET R_HOME=/GCModeller/src/R-sharp/App/net6.0-windows
SET R_ENV="%R_HOME%/R#.exe"

%R_ENV% ./build-githubpage.R

pause
313 changes: 313 additions & 0 deletions vignettes/MLkit/CNN.html
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,313 @@
<!DOCTYPE html><html lang="zh-CN">
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8" />
<meta http-equiv="X-UA-Compatible" content="IE=Edge" />
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, minimum-scale=1.0, maximum-scale=1.0" />
<title>CNN</title>
<meta name="author" content="[email protected]" />
<meta name="copyright" content="SMRUCC genomics Copyright (c) 2022" />
<meta name="keywords" content="R#; CNN; MLkit" />
<meta name="generator" content="https://github.com/rsharp-lang" />
<meta name="theme-color" content="#333" />
<meta name="description" content="feed-forward phase of deep Convolutional Neural Networks..." />
<meta class="foundation-data-attribute-namespace" />
<meta class="foundation-mq-xxlarge" />
<meta class="foundation-mq-xlarge" />
<meta class="foundation-mq-large" />
<meta class="foundation-mq-medium" />
<meta class="foundation-mq-small" />
<meta class="foundation-mq-topbar" />
<style>

.table-three-line {
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) */
border:solid #000000; /* 设置边框属性；样式(solid=实线)、颜色(#999=灰) */
border-width:2px 0 2px 0px; /* 设置边框状粗细：上 右 下 左 = 对应：1px 0 0 1px */
}
.left-1{
border:solid #000000;border-width:1px 1px 2px 0px;padding:2px;
font-weight:bolder;
}
.right-1{
border:solid #000000;border-width:1px 0px 2px 1px;padding:2px;
font-weight:bolder;
}
.mid-1{
border:solid #000000;border-width:1px 1px 2px 1px;padding:2px;
font-weight:bolder;
}
.left{
border:solid #000000;border-width:1px 1px 1px 0px;padding:2px;
}
.right{
border:solid #000000;border-width:1px 0px 1px 1px;padding:2px;
}
.mid{
border:solid #000000;border-width:1px 1px 1px 1px;padding:2px;
}
table caption {font-size:14px;font-weight:bolder;}
</style>
</head>
<body>
<table width="100%" summary="page for {CNN}">
<tbody>
<tr>
<td>{CNN}</td>
<td style="text-align: right;">R# Documentation</td>
</tr>
</tbody>
</table>
<h1>CNN</h1>
<hr />
<p style=" font-size: 1.125em; line-height: .8em; margin-left: 0.5%; background-color: #fbfbfb; padding: 24px; ">
<code>
<span style="color: blue;">require</span>(<span style="color: black; font-weight: bold;">R</span>);
<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>;
</code>
</p>
<p><p>feed-forward phase of deep Convolutional Neural Networks</p></p>
<blockquote>
<p style="font-style: italic; font-size: 0.9em;">
<p>feed-forward phase of deep Convolutional Neural Networks</p>
</p>
</blockquote>
<div id="main-wrapper">
<table class="table-three-line">
<tbody><tr>
<td id="n_threads">
<a href="./CNN/n_threads.html">n_threads</a>
</td>
<td><p>get/set of the CNN parallel thread number</p></td>
</tr>
<tr>
<td id="cnn">
<a href="./CNN/cnn.html">cnn</a>
</td>
<td><p>Create a new CNN model</p>

<p>Convolutional neural network (CNN) is a regularized type of feed-forward<br />
neural network that learns feature engineering by itself via filters <br />
(or kernel) optimization. Vanishing gradients and exploding gradients, <br />
seen during backpropagation in earlier neural networks, are prevented by <br />
using regularized weights over fewer connections.</p></td>
</tr>
<tr>
<td id="input_layer">
<a href="./CNN/input_layer.html">input_layer</a>
</td>
<td><p>The input layer is a simple layer that will pass the data though and<br />
create a window into the full training data set. So for instance if<br />
we have an image of size 28x28x1 which means that we have 28 pixels<br />
in the x axle and 28 pixels in the y axle and one color (gray scale),<br />
then this layer might give you a window of another size example 24x24x1<br />
that is randomly chosen in order to create some distortion into the<br />
dataset so the algorithm don't over-fit the training.</p></td>
</tr>
<tr>
<td id="regression_layer">
<a href="./CNN/regression_layer.html">regression_layer</a>
</td>
<td></td>
</tr>
<tr>
<td id="conv_layer">
<a href="./CNN/conv_layer.html">conv_layer</a>
</td>
<td><p>This layer uses different filters to find attributes of the data that<br />
affects the result. As an example there could be a filter to find<br />
horizontal edges in an image.</p></td>
</tr>
<tr>
<td id="conv_transpose_layer">
<a href="./CNN/conv_transpose_layer.html">conv_transpose_layer</a>
</td>
<td></td>
</tr>
<tr>
<td id="lrn_layer">
<a href="./CNN/lrn_layer.html">lrn_layer</a>
</td>
<td><p>This layer is useful when we are dealing with ReLU neurons. Why is that?<br />
Because ReLU neurons have unbounded activations and we need LRN to normalize<br />
that. We want to detect high frequency features with a large response. If we<br />
normalize around the local neighborhood of the excited neuron, it becomes even<br />
more sensitive as compared to its neighbors.</p>

<p>At the same time, it will dampen the responses that are uniformly large in any<br />
given local neighborhood. If all the values are large, then normalizing those<br />
values will diminish all of them. So basically we want to encourage some kind<br />
of inhibition and boost the neurons with relatively larger activations. This<br />
has been discussed nicely in Section 3.3 of the original paper by Krizhevsky et al.</p></td>
</tr>
<tr>
<td id="tanh_layer">
<a href="./CNN/tanh_layer.html">tanh_layer</a>
</td>
<td><p>Implements Tanh nonlinearity elementwise x to tanh(x)<br />
so the output is between -1 and 1.</p></td>
</tr>
<tr>
<td id="softmax_layer">
<a href="./CNN/softmax_layer.html">softmax_layer</a>
</td>
<td><p>[*loss_layers] This layer will squash the result of the activations in the fully<br />
connected layer and give you a value of 0 to 1 for all output activations.</p></td>
</tr>
<tr>
<td id="relu_layer">
<a href="./CNN/relu_layer.html">relu_layer</a>
</td>
<td><p>This is a layer of neurons that applies the non-saturating activation<br />
function f(x)=max(0,x). It increases the nonlinear properties of the<br />
decision function and of the overall network without affecting the<br />
receptive fields of the convolution layer.</p></td>
</tr>
<tr>
<td id="leaky_relu_layer">
<a href="./CNN/leaky_relu_layer.html">leaky_relu_layer</a>
</td>
<td></td>
</tr>
<tr>
<td id="maxout_layer">
<a href="./CNN/maxout_layer.html">maxout_layer</a>
</td>
<td><p>Implements Maxout nonlinearity that computes x to max(x)<br />
where x is a vector of size group_size. Ideally of course,<br />
the input size should be exactly divisible by group_size</p></td>
</tr>
<tr>
<td id="sigmoid_layer">
<a href="./CNN/sigmoid_layer.html">sigmoid_layer</a>
</td>
<td><p>Implements Sigmoid nonlinearity elementwise x to 1/(1+e^(-x))<br />
so the output is between 0 and 1.</p></td>
</tr>
<tr>
<td id="pool_layer">
<a href="./CNN/pool_layer.html">pool_layer</a>
</td>
<td><p>This layer will reduce the dataset by creating a smaller zoomed out<br />
version. In essence you take a cluster of pixels take the sum of them<br />
and put the result in the reduced position of the new image.</p></td>
</tr>
<tr>
<td id="dropout_layer">
<a href="./CNN/dropout_layer.html">dropout_layer</a>
</td>
<td><p>This layer will remove some random activations in order to<br />
defeat over-fitting.</p></td>
</tr>
<tr>
<td id="full_connected_layer">
<a href="./CNN/full_connected_layer.html">full_connected_layer</a>
</td>
<td><p>Neurons in a fully connected layer have full connections to all<br />
activations in the previous layer, as seen in regular Neural Networks.<br />
Their activations can hence be computed with a matrix multiplication<br />
followed by a bias offset.</p></td>
</tr>
<tr>
<td id="gaussian_layer">
<a href="./CNN/gaussian_layer.html">gaussian_layer</a>
</td>
<td></td>
</tr>
<tr>
<td id="sample_dataset">
<a href="./CNN/sample_dataset.html">sample_dataset</a>
</td>
<td></td>
</tr>
<tr>
<td id="sample_dataset.image">
<a href="./CNN/sample_dataset.image.html">sample_dataset.image</a>
</td>
<td></td>
</tr>
<tr>
<td id="auto_encoder">
<a href="./CNN/auto_encoder.html">auto_encoder</a>
</td>
<td></td>
</tr>
<tr>
<td id="training">
<a href="./CNN/training.html">training</a>
</td>
<td><p>Do CNN network model training</p></td>
</tr>
<tr>
<td id="ada_delta">
<a href="./CNN/ada_delta.html">ada_delta</a>
</td>
<td><p>Adaptive delta will look at the differences between the expected result and the current result to train the network.</p></td>
</tr>
<tr>
<td id="ada_grad">
<a href="./CNN/ada_grad.html">ada_grad</a>
</td>
<td><p>The adaptive gradient trainer will over time sum up the square of<br />
the gradient and use it to change the weights.</p></td>
</tr>
<tr>
<td id="adam">
<a href="./CNN/adam.html">adam</a>
</td>
<td><p>Adaptive Moment Estimation is an update to RMSProp optimizer. In this running average of both the<br />
gradients and their magnitudes are used.</p></td>
</tr>
<tr>
<td id="nesterov">
<a href="./CNN/nesterov.html">nesterov</a>
</td>
<td><p>Another extension of gradient descent is due to Yurii Nesterov from 1983,[7] and has been subsequently generalized</p></td>
</tr>
<tr>
<td id="sgd">
<a href="./CNN/sgd.html">sgd</a>
</td>
<td><p>Stochastic gradient descent (often shortened in SGD), also known as incremental gradient descent, is a<br />
stochastic approximation of the gradient descent optimization method for minimizing an objective function<br />
that is written as a sum of differentiable functions. In other words, SGD tries to find minimums or<br />
maximums by iteration.</p></td>
</tr>
<tr>
<td id="window_grad">
<a href="./CNN/window_grad.html">window_grad</a>
</td>
<td><p>This is AdaGrad but with a moving window weighted average<br />
so the gradient is not accumulated over the entire history of the run.<br />
it's also referred to as Idea #1 in Zeiler paper on AdaDelta.</p></td>
</tr>
<tr>
<td id="predict">
<a href="./CNN/predict.html">predict</a>
</td>
<td></td>
</tr>
<tr>
<td id="CeNiN">
<a href="./CNN/CeNiN.html">CeNiN</a>
</td>
<td><p>load a CNN model from file</p></td>
</tr>
<tr>
<td id="detectObject">
<a href="./CNN/detectObject.html">detectObject</a>
</td>
<td><p>classify a object from a given image data</p></td>
</tr>
<tr>
<td id="saveModel">
<a href="./CNN/saveModel.html">saveModel</a>
</td>
<td><p>save the CNN model into a binary data file</p></td>
</tr></tbody>
</table>
</div>
<hr />
<div style="text-align: center;">[<a href="../index.html">Document Index</a>]</div>
</body>
</html>
Loading

0 comments on commit e0ff8e7

Please sign in to comment.