TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

A Beginner's Guide to Understanding Convolutional Neural Networks

367 pointsby kilimchoialmost 9 years ago

8 comments

Dzugarualmost 9 years ago
Have yet to see an illustration that grasps multichannel convolution filters (MCCF) concept clearly. Why those channel stack size keep growing? How are they actually connected?<p>The thing that each conv filter consists of kernels in multiple channels (that&#x27;s why first layer filter visualisations are colored btw - color image is a &quot;3-dimensional&quot; image) - and we convolve each kernel with corresponding input channel, then <i>sum</i> (that&#x27;s the key) the responses. Then having multiple MCCF (usually more at each layer) yields a new multi-channel image (say, 16 channels) and we apply new set of (say, 32) 16-channeled MCCFs to it (which we cannot visualise by themselves anymore, we need a 16-dimensional image for each filter) yielding 32-channel image. That sort of thing is almost never explained properly.
评论 #12137376 未加载
评论 #12141542 未加载
评论 #12135991 未加载
评论 #12135992 未加载
chrisrukalmost 9 years ago
<a href="http:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;1602.04105#" rel="nofollow">http:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;1602.04105#</a> -- This paper is awesome for a use of CNNs, for automatic modulation recognition of RF signals.<p>I&#x27;m attempting to use their approach with GNU Radio currently -<p><a href="https:&#x2F;&#x2F;radioml.com&#x2F;blog&#x2F;2016&#x2F;07&#x2F;18&#x2F;towards-a-gnu-radio-cnn-tensorflow-block&#x2F;" rel="nofollow">https:&#x2F;&#x2F;radioml.com&#x2F;blog&#x2F;2016&#x2F;07&#x2F;18&#x2F;towards-a-gnu-radio-cnn-...</a>
danielmorozoffalmost 9 years ago
Great writeup from Stanford CS231 course: <a href="http:&#x2F;&#x2F;cs231n.github.io&#x2F;convolutional-networks&#x2F;" rel="nofollow">http:&#x2F;&#x2F;cs231n.github.io&#x2F;convolutional-networks&#x2F;</a>
sjnair96almost 9 years ago
Damn the author is a freshman!
thallukrishalmost 9 years ago
A human child learns much more easily by seeing only a handful of images of a cat and then almost being able to say any type of cat image as it grows (without ever seeing 1 million or billion images). So, there seem to be something that shows that more than the amount of data, the &quot;reality&quot; of seeing a real cat probably includes all possible aspects of a Cat ? There seem to be something missing with this whole deep learning stuff and the way it is trying to simulate the human cognition.
vonnikalmost 9 years ago
Here&#x27;s an intro to ConvNets in Java: <a href="http:&#x2F;&#x2F;deeplearning4j.org&#x2F;convolutionalnets.html" rel="nofollow">http:&#x2F;&#x2F;deeplearning4j.org&#x2F;convolutionalnets.html</a><p>Karpathy&#x27;s stuff is also great: <a href="https:&#x2F;&#x2F;cs231n.github.io&#x2F;" rel="nofollow">https:&#x2F;&#x2F;cs231n.github.io&#x2F;</a>
crncostaalmost 9 years ago
Very well illustrated.
cynicaldevilalmost 9 years ago
I am new to CNNs&#x2F;machine learning, but here&#x27;s my $0.02: Regardless of which technique you use, it seems that the amount of data required to learn is too high. This article talks about neural networks accessing billions of photographs, a number which is nowhere near the number of photos&#x2F;objects&#x2F;whatever a human sees in a lifetime. Which leads me to the conclusion that we aren&#x27;t extracting much information from the data. These techniques aren&#x27;t able to calculate how the same object might look under different lighting conditions, different viewing angles, positions, sizes, and so on. Instead, companies just use millions of images to &#x27;encode&#x27; the variations into their networks.<p>Imo there should be a push towards adapting CNNs to calculate&#x2F;predict how the object might look under different conditions, which might lead to other improvements. This could also be extended to areas other than image recognition.
评论 #12136259 未加载
评论 #12136451 未加载