KI-Glossar
Das vollständige Wörterbuch der Künstlichen Intelligenz
Principal Component Analysis
Linear technique that projects data onto axes of maximum variance to reduce dimensionality.
t-SNE
Non-linear algorithm that preserves local structures by transforming similarities into probabilities.
UMAP
Non-linear dimensionality reduction technique based on algebraic topology and manifold learning.
Autoencoders
Neural network architecture that learns to compress and reconstruct data to reduce their dimensionality.
Independent Component Analysis
Statistical method that separates a multivariate signal into underlying independent components.
Linear Discriminant Analysis
Supervised technique that maximizes separability between classes while reducing dimensionality.
Matrix Factorization
Decomposition of a matrix into a product of lower-rank matrices to reveal latent structures.
ISOMAP
Non-linear algorithm that preserves geodesic distances on an embedded manifold.
Locally Linear Embedding
Technique that preserves local neighborhood relationships in the reduced dimension space.
Multidimensional Scaling
Method that represents objects in a lower-dimensional space while preserving their mutual distances.
Non-negative Matrix Factorization
Matrix factorization with non-negativity constraint producing additive representations.
Random Projection
Simple technique that projects data onto a random subspace while preserving distances.
Diffusion Maps
Method based on diffusion processes to capture the intrinsic geometry of data.
Feature Selection
Selection of an optimal subset of original features rather than projection onto new dimensions.
Kernel PCA
PCA extension using kernel functions to capture non-linear relationships in the data.