Glossario IA
Il dizionario completo dell'Intelligenza Artificiale
Generator
Neural network in a GAN responsible for creating new data by learning to map latent space to training data space to generate realistic samples.
Discriminator
Classifier neural network in a GAN that distinguishes between authentic samples and artificially generated ones, providing a learning signal to the generator.
Autoencoder GAN
GAN variant incorporating an autoencoder structure where the generator acts as an encoder-decoder learning to reconstruct normal data and identify anomalies.
Normality Distribution
Statistical distribution of normal data learned by the GAN serving as a reference to identify observations that significantly deviate from it.
Threshold-based Detection
Anomaly classification method using a predefined threshold on anomaly scores to separate normal observations from abnormal ones.
Deep Feature Learning
Automatic learning of complex hierarchical representations by GAN's deep networks to capture the subtleties of normal data.
Discriminative Features
Features extracted by the discriminator allowing fine distinction between normal and abnormal data for precise detection.