<p>In <a href="https://github.com/mlpack/mlpack/pull/757#discussion_r75576884">src/mlpack/methods/ann/layer/inception_layer.hpp</a>:</p>
<pre style='color:#555'>&gt; +    // Forward pass for 3x3 pool path. 
&gt; +    pool3.InputParameter() = input;
&gt; +    pool3.Forward(input, pool3.OutputParameter());
&gt; +    Pad(pool3.OutputParameter(), 1, 1, convPool.InputParameter());
&gt; +    convPool.Forward(pool3.OutputParameter(), convPool.OutputParameter());
&gt; +    biasPool.Forward(convPool.OutputParameter(), biasPool.OutputParameter());
&gt; +    basePool.InputParameter() = biasPool.OutputParameter();
&gt; +    basePool.Forward(convPool.OutputParameter(), basePool.OutputParameter());
&gt; +
&gt; +    // concatenate outputs of all the paths.
&gt; +    output = arma::join_slices( 
&gt; +              arma::join_slices(
&gt; +                arma::join_slices( 
&gt; +                  base1.OutputParameter(), base3.OutputParameter() ), 
&gt; +                  base5.OutputParameter() ), basePool.OutputParameter());
&gt; +
</pre>
<p>As you already pointed out the inception layer is basically a container for a couple of convolution and pooling layer. Since they are abstracted through the inception layer, the main network has now knowledge about the containing layer. So, the inception layer has to forward any weights to any capsulated layer.</p>

<p style="font-size:small;-webkit-text-size-adjust:none;color:#666;">&mdash;<br />You are receiving this because you are subscribed to this thread.<br />Reply to this email directly, <a href="https://github.com/mlpack/mlpack/pull/757/files/06d923321f246f2c6ead9ad56e99309fe24a6f5c#r75576884">view it on GitHub</a>, or <a href="https://github.com/notifications/unsubscribe-auth/AJ4bFFSNWCVU27PI_P1ffDHk-yboky_Sks5qhuCLgaJpZM4JejWj">mute the thread</a>.<img alt="" height="1" src="https://github.com/notifications/beacon/AJ4bFJqO82lB0bvqyYr5PX4ZTMOVOoPYks5qhuCLgaJpZM4JejWj.gif" width="1" /></p>
<div itemscope itemtype="http://schema.org/EmailMessage">
<div itemprop="action" itemscope itemtype="http://schema.org/ViewAction">
  <link itemprop="url" href="https://github.com/mlpack/mlpack/pull/757/files/06d923321f246f2c6ead9ad56e99309fe24a6f5c#r75576884"></link>
  <meta itemprop="name" content="View Pull Request"></meta>
</div>
<meta itemprop="description" content="View this Pull Request on GitHub"></meta>
</div>

<script type="application/json" data-scope="inboxmarkup">{"api_version":"1.0","publisher":{"api_key":"05dde50f1d1a384dd78767c55493e4bb","name":"GitHub"},"entity":{"external_key":"github/mlpack/mlpack","title":"mlpack/mlpack","subtitle":"GitHub repository","main_image_url":"https://cloud.githubusercontent.com/assets/143418/17495839/a5054eac-5d88-11e6-95fc-7290892c7bb5.png","avatar_image_url":"https://cloud.githubusercontent.com/assets/143418/15842166/7c72db34-2c0b-11e6-9aed-b52498112777.png","action":{"name":"Open in GitHub","url":"https://github.com/mlpack/mlpack"}},"updates":{"snippets":[{"icon":"PERSON","message":"@zoq in #757: As you already pointed out the inception layer is basically a container for a couple of convolution and pooling layer. Since they are abstracted through the inception layer, the main network has now knowledge about the containing layer. So, the inception layer has to forward any weights to any capsulated layer.\r\n"}],"action":{"name":"View Pull Request","url":"https://github.com/mlpack/mlpack/pull/757/files/06d923321f246f2c6ead9ad56e99309fe24a6f5c#r75576884"}}}</script>