Convolution vs Cross-Correlation. How most CNNs do not compute convolutions. | ❓ #Shorts

preview_player
Показать описание
Most CNNs do not compute convolutions, but cross-correlations. This is how, in 60 seconds.
Watch and like it to support us! only the top 100 most viewed videos qualify for the contest. ✌

Discussed AI Coffee Break Quiz questions:

Read more here:

Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏
donor, Dres. Trost GbR

▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀

🔗 Links:

#AICoffeeBreak #AICoffeeBreakQuiz #MsCoffeeBean #MachineLearning #AI
Рекомендации по теме
Комментарии
Автор

Do libraries do this just to punish and confuse us

sadface
Автор

So the pytorch implementations - do they have flipped kernels or normal? Also, does it matter?

abhilashnandy
Автор

I always found this discussion a bit pedantic. Whether the operation is implemented as cross-correlation or convolution ultimately doesn't really matter, does it? The same is learned, only the index of the parameter is different.

majork.
Автор

Hey, will you ever cover spiking neural nets or STDP's for audio and visual recognition? Thanks!

neocephalon
Автор

Hi, are pytorch and tensorflow implementation compatible? (are both crosscorrelations?)

piotr
Автор

Thanks for this eye-opener ma'am. I have seen multiple of your comments on NYU21 yann's playlist, especially on EBM videos. It would be really helpful to see your intuition on energy based models. Could you please make a video on it. 😅😅 That would be really great. Thank you.

siddhantrai