Videos

Convolutional Neural Networks (CNNs) explained



deeplizard

CNNs for deep learning. Blog for this vid! http://deeplizard.com/learn/video/YRhxdVk_sIs

#21 in Machine Leaning / Deep Learning for Programmers Playlist
https://www.youtube.com/playlist?list=PLZbbT5o_s2xq7LwI2y8_QtvuXZedL6tQU

In this video, we explain the concept of convolutional neural networks, how they’re used, and how they work on a technical level. We also discuss the details behind convolutional layers and filters.

fast.ai lesson 4:
http://course.fast.ai/lessons/lesson4.html

Follow deeplizard on Twitter:
https://twitter.com/deeplizard

Follow deeplizard on Steemit:
https://steemit.com/@deeplizard

Become a patron:
https://www.patreon.com/deeplizard

Support deeplizard:
Bitcoin: 1AFgm3fLTiG5pNPgnfkKdsktgxLCMYpxCN
Litecoin: LTZ2AUGpDmFm85y89PFFvVR5QmfX6Rfzg3
Ether: 0x9105cd0ecbc921ad19f6d5f9dd249735da8269ef

Recommended books:
The Most Human Human: What Artificial Intelligence Teaches Us About Being Alive: http://amzn.to/2GtjKqu

Source

Similar Posts

43 thoughts on “Convolutional Neural Networks (CNNs) explained
  1. Halo,
    So in the example that mentioned in this video, a filter is used and convolved over the image "7" and recreated an output of "7" by doing dot product…..So my question is that whether this is similar to autoencoders?….because, in both a recreation of the input at the output is happening…right?

  2. Halo,
    I would like to know regarding flattening in deep learning, and also………
    I would like to know how to create a new image dataset ?…….whether we should concern about image resolution and other criteria while downloading and saving…..etc….

  3. Hey deeplizard,

    thanks for the great video (and series)! I perceive your explanations as very clear and audience-focused. I really enjoy your videos! 🙂

    I was wondering how exactly convolution works. You said that the convolution operation you perform is based on a dot product between the filter and a given patch of the data. I tried to simulate this process and it seems to me that to get the same results as in your excel simulation one should rather perform element-wise multiplication and then sum all the elements of the resulting matrix. How do you think about it?

    Thanks again for your effort and great content! <3

  4. The Excel example was beautiful! Quick question: does the size of the image that is passed decrease over each layer? I didn't count the rows and columns of the dot product result, but intuitively it seems that it will decrease?

  5. we don't dot product the filter with a region on input. it's an element wise multiplication , each cell times the corresponding cell and then we sum them all to get one output.

  6. I really like the way you explain. No matter if you're new to this topic or just refreshing from a different POV. Subscribed and thanks!

Comments are closed.

WP2Social Auto Publish Powered By : XYZScripts.com