KI-Glossar
Das vollständige Wörterbuch der Künstlichen Intelligenz
Convolutional VAE
Variant of the Variational Autoencoder incorporating convolutional layers to effectively capture spatial and hierarchical structures in image data.
Convolutional Encoder
Convolutional neural network that transforms an input image into distribution parameters (mean and variance) in the latent space.
Convolutional Decoder
Deconvolution or transposed architecture that reconstructs an image from a sampled latent vector.
ELBO Loss Function
Evidence Lower BOund optimized in VAEs, combining reconstruction error and KL-divergence to balance fidelity and regularization.
Transposed Convolution Architecture
Decoder structure using transposed convolutions (deconvolutions) to upsample and reconstruct images.
Variational Inference Mechanism
Process of approximating a complex posterior by a simpler and tractable parametric distribution.
Conditional Image Generation
CVAE extension enabling generation of images conditioned on specific attributes or labels.
KL-Divergence Regularization
Penalty term in the loss function forcing the latent distribution to follow a standard Gaussian.
Deep Convolutional Blocks
Stacking of convolutional layers, pooling, and activation to extract hierarchical features.
Transfer Learning with CVAE
Adaptation of a pre-trained CVAE for classification or segmentation tasks with minimal fine-tuning.