AI 용어집
인공지능 완전 사전
Classical Autoencoders
Fundamental autoencoder architecture with symmetric encoder-decoder for data compression and reconstruction.
Variational Autoencoders (VAE)
Probabilistic autoencoders that generate data by learning a latent distribution for creating new samples.
Denoising Autoencoders
Autoencoders trained to reconstruct clean data from inputs corrupted by random noise.
Sparse Autoencoders
Autoencoders using a sparsity constraint on the activation of hidden neurons for efficient representation.
Contractive Autoencoders
Autoencoders penalizing the sensitivity of the representation to input variations for better robustness.
Convolutional Autoencoders
Autoencoders using convolutional layers to efficiently process image and spatial data.
Deep Autoencoders
Autoencoders with multiple hidden layers enabling complex hierarchical nonlinear compression.
Recurrent Autoencoders
Autoencoders based on recurrent networks for sequential and temporal data processing.
Adversarial Autoencoders
Autoencoders combining adversarial learning to improve the quality of latent representations.
Principal Component Analysis (PCA)
Classical linear method for dimensionality reduction that maximizes the variance preserved in the projected data.
t-SNE and UMAP
Non-linear algorithms for visualization and dimensionality reduction preserving the local structure of data.
Memory Autoencoders
Autoencoders incorporating memory mechanisms to efficiently store and retrieve complex patterns.
Autoencoders for Time Series
Specialized architectures for compression and analysis of temporal data with sequential dependencies.
Factor Analysis
Statistical technique identifying underlying latent factors that explain correlations between variables.
Variable Selection Reduction
Methods that eliminate or select the most relevant features to reduce dimensionality.