Can you imagine creating new photos that look like old ones?
Autoencoder
Can you imagine creating new photos that look like old ones?
Ever tried to fill in missing parts of a blurry family photo?
Imagine a smart program that can take a blurry photo and fill in the missing details. It learns to do this by practicing with lots of photos, getting better at guessing the missing parts.
Example
If you show the program a blurry photo of a cat, it might fill in the missing details so well that it looks like a clear photo of a cat.
Remember this
A Variational Autoencoder (VAE) learns to fill in missing parts of photos, creating clearer versions from blurry ones.
Text adapted from Wikipedia, licensed under CC BY-SA 4.0.
Related concepts
Artificial intelligence in music
Can AI compose music like a human?
Unsupervised learning
How can we find hidden patterns in massive, unlabeled data?
data augmentation does for generalization: artificially expands training set
How can you teach a computer to see better?
autoencoders learn the data manifold
Autoencoders compress data manifold by forcing information through a bottleneck layer, learning efficient representations
batch size affects generalization: larger batches find sharper minima
Larger batch sizes lead to sharper minima, enhancing generalization by providing more accurate gradient estimates
Reparameterization trick
Reparameterization trick enables differentiable sampling for VAE training
Swipe through more Machine Learning concepts
Open Pocket Polymath