AI-ordlista
Den kompletta ordlistan över AI
Variational Auto-Encoder
Generative neural network architecture that learns a probabilistic distribution in the latent space rather than a deterministic encoding, enabling the generation of new data through sampling.
Probabilistic Encoder
Neural network that maps input data to the parameters (mean and variance) of a distribution in the latent space rather than a fixed point.
Generative Decoder
Neural network that reconstructs the original data from samples in the latent space, modeling the conditional probability p(x|z).
Inference Posterior
Approximate distribution q(z|x) learned by the encoder to estimate the true posterior p(z|x) which is generally intractable.
Reparameterization Noise
Random variable ε ~ N(0, I) introduced during the reparameterization trick to enable differentiation of the sampling operation.
Continuous Space
Property of the VAE latent space where neighboring points correspond to similar data, enabling interpolation and semantic navigation.
Latent Sampling
Process of generating new samples by drawing random points from the learned distribution in the latent space and passing them through the decoder.
Deterministic Auto-Encoder
Classical variant where encoding produces a unique vector in the latent space without a stochastic component, unlike the VAE.
Architecture Variationnelle
Structure neuronale intégrant des couches probabilistes et des mécanismes d'échantillonnage pour modéliser des distributions plutôt que des mappings déterministes.