Adversarial Autoencoders: The Reconstruction Loss
Beginner
The problem, from first principles
An adversarial autoencoder (Makhzani et al., 2015) is an autoencoder whose latent codes are also pushed toward a chosen prior by a discriminator. The first part is an ordinary reconstruction loss, mean squared error here.
From theory to code
Implement reconstruction_mse(x, x_hat), the mean squared error over all entries.
Constraints
- Average over every element.
Hints
Hint 1
Square the differences, then take the mean.
Keywords
research-papersunsupervisedautoencodersadversarial
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.