Generative Adversarial Networks

Рет қаралды 7,582

Leo Isikdogan

Күн бұрын

Пікірлер

@mr.chanshu1610 2 жыл бұрын

Can anyone explain this part 3:17

@DerIntergalaktische 6 жыл бұрын

Great video series! How does a generator network look like? I have read that you have to use fractional stride convolutions, but I can't find any good explanation on how such convolutions are done.

@leoisikdogan 6 жыл бұрын

Thanks! How a generator network looks like depends on the nature of the task. For image-to-image tasks, it's typically an autoencoder-like network. The decoder may use fractional stride convolutions to upsample the input. You can check out my video on unsupervised learning to learn more about autoencoders and my video on convolutional netural networks to see how fractional strides work.

@geogeo14000 4 жыл бұрын

Great video again. I'm new to AI and sorry for my naive question, but I don't really get the moment you talk about feedforward CNN. I thought that in a feedforward model like perceptrons, there was no backpropagation instead in CNN or recurrent NN there was. Or convolutional does it only mean what u explained in yr video on this topic, i.e. a*b ? I feel I am missing sthg and I'm messing too so XD Anyway thanx again for your videos, already shared on my fb account, as others said it u deserve a much larger audience ;)

@leoisikdogan 4 жыл бұрын

Thanks. CNNs are trained the same way perceptrons are trained. Except for some gradient-free methods, nearly all neural networks are trained using backpropagation. This includes perceptrons, feedforward CNNs, RNNs, and so on.

@sainath66666 5 жыл бұрын

Please do a few livestream!!!

@leoisikdogan 5 жыл бұрын

Thanks for your suggestion. Livestreaming is not feasible for my channel at the moment, because I don't have a large enough audience and it's hard to find a time that suits my followers living in different time zones across the world.