Convolution Padding - Neural Networks

Поделиться
HTML-код
  • Опубликовано: 3 ноя 2024

Комментарии • 19

  • @Anodder1
    @Anodder1 2 года назад +8

    I am working with neural networks for years now and every now and then I come back to the fundamentals to check my current understanding. Your videos are truly great! They are short, accurate, visually helpful and contain practical knowledge. Thank you very much for making these and I hope you get to make a full series!

  • @mrcallum1995
    @mrcallum1995 Год назад +4

    I just watched all your videos. Didn’t know anything about CNNs before, now I feel I have a solid foundation. Please make some more!

  • @Sydra.
    @Sydra. Год назад +1

    Underrated channel. You should have much more views.

  • @charlieng3347
    @charlieng3347 8 месяцев назад

    Thank you so much for this video series. It really helped me with understanding convolution neural network.

  • @DanielTorres-gd2uf
    @DanielTorres-gd2uf Год назад

    Keep it up! These are amazing, straightforward technical videos that are invaluable for someone starting out.

  • @rubiczhang5593
    @rubiczhang5593 5 месяцев назад

    That's a really good job, you have save me from struggling with the AI. Thank you from China

  • @nojoodal-ghamdi5579
    @nojoodal-ghamdi5579 Год назад

    I just shared your channel with my AI professor and told her to include those videos within our course because it made everything clear! Thank you

  • @grayboywilliams
    @grayboywilliams Год назад +1

    Man your videos are revolutionary. Please do a set on transformers next, I have spent many hours trying to visualize them!

  • @_general_error
    @_general_error Год назад +2

    Implementing padding in convolution would slow the whole algortihm down on architectures like the TPU. It also applies to most GPUs, you simply want to avoid branched code. It gives you huge performance benefits to prepare the data in memory and stream it through the computing system.

  • @davidalexander7994
    @davidalexander7994 2 года назад

    I'm new to NN and these are really helpful 👍

  • @pabitrasahoo164
    @pabitrasahoo164 2 года назад

    Great explanation

  • @Sarkermiraz
    @Sarkermiraz Год назад

    Great explanation. Please make a video on 3d convolution and strides.

  • @xl0xl0xl0
    @xl0xl0xl0 Год назад

    I love your visualization. I always end up drawing blocks when I try to understand how the data flows in an architecture, but your visualizations are to scale, have colors, and give a very good idea how the "volumes" of the data and parameters change throughout the model. It's not just an aid for understanding basic concepts like convolution - it's a bit of an overkill for that, but it's excellent for visualizing a large NNs. I wonder if it would be possible to create a "model explorer" that would allow you to explore interactively in 3D, like a video game. Maybe even with VR? But just 3D is probably plenty good.

  • @126sivgucsivanshgupta2
    @126sivgucsivanshgupta2 2 года назад

    Will you be explaining the training process ? ie how gradient decent works in CNN's
    Also will you be explaining Max pooling?

  • @Henriiyy
    @Henriiyy Год назад

    "all the way back to convolutions origin in signal processing"? Looking quickly in the "history" section on wikipedia could tell you, that convolution was first invented by d'Alembert in 1754...

  • @leamon9024
    @leamon9024 2 года назад

    Amazing work! Could you also make videos about vision transformer? Thanks

    • @animatedai
      @animatedai  2 года назад +3

      After the current series on convolution, my tentative plans are to make a series on attention, and I could make vision transformers part of that series. Thanks for the suggestion.

    • @leamon9024
      @leamon9024 2 года назад

      @@animatedai Cool! Would love to see attention related animation tutorial as well!!

  • @omridrori3286
    @omridrori3286 Год назад

    Can you upload one on normalization techniques in convolutional neural network ?
    Or better on transformers in vision it is hard to get these concepts