Słownik AI
Kompletny słownik sztucznej inteligencji
Adversarial Transfer Learning
Transfer learning approach using generative adversarial networks to reduce the distribution gap between source and target domains.
Distribution Discrepancy
Quantitative measure of the difference between probability distributions of source and target domain data.
Gradient Reversal Layer
Neural network layer that inverts the gradient during backpropagation to facilitate learning of domain-invariant features.
Domain Discriminator
Neural network trained to distinguish features from the source domain from those of the target domain in an adversarial framework.
Source Domain
Original domain containing labeled data used for the initial training of the model before transfer.
Target Domain
Destination domain where the model should perform, typically with few or no labeled data available.
Adversarial Training
Simultaneous training process of two competing networks, where a generator attempts to deceive a discriminator.
Feature Extractor
Component of the neural network responsible for transforming raw data into abstract feature representations.
Domain Classifier
Binary classifier trained to identify if a feature comes from the source or target domain.