Glossario IA

📖

termini

Stochastic bottleneck

Intermediate layer of VAEs where information is compressed into random variables following probability distributions, unlike the deterministic bottleneck of classical autoencoders.

📖

termini

Latent sampling

Process of sampling points in the latent space according to the distributions learned by the encoder, enabling the generation of diverse and realistic new data.

📖

termini

VAE loss function

Composite objective function combining the reconstruction loss (typically mean squared error or binary cross-entropy) and the KL divergence regularization term.

📖

termini

Stochastic regularization

Inherent mechanism in VAEs that prevents overfitting by forcing latent distributions to remain close to the prior distribution, ensuring a continuous and interpolable latent space.