AI-ordlista
Den kompletta ordlistan över AI
Standard VAE Architecture
Fundamental structure of variational autoencoders with encoder, decoder and probabilistic latent space.
Latent Space and Distribution
Management of latent variables following probabilistic distributions to capture data structure.
ELBO Loss Function
Evidence Lower Bound optimizing reconstruction while regularizing the latent space via KL divergence.
Conditional VAEs
Variational autoencoders integrating conditions to generate controlled data based on specific attributes.
Hierarchical VAEs
Multi-level architectures capturing abstractions at different scales in the latent space.
Convolutional VAEs
Application of VAEs to image data using convolutional layers to capture spatial structures.
Recurrent VAEs
Extension of VAEs to sequential data such as text or time series with recurrent mechanisms.
Posterior Collapse
Phenomenon where the encoder ignores the input data and its solutions through architectural modifications.
β-VAE and Variants
Modifications of the regularization term to control the trade-off between reconstruction and latent space capacity.
Representative Disentanglement
VAE techniques to learn independent and interpretable latent factors in data.
Amortized Variational Inference
Learning process of a neural encoder to efficiently approximate the posterior distribution.
Semi-Supervised VAEs
Combination of supervised and unsupervised learning using VAEs with partially labeled data.
VAEs for Imputation
Using VAEs to handle and impute missing values in incomplete datasets.
Adversarial VAEs
Fusion of VAEs with GANs to improve the quality of generated samples while retaining inference.
Normalizing Flows in VAEs
Use of bijective transformations to enrich the family of approximate posterior distributions.