Glosarium AI
Kamus lengkap Kecerdasan Buatan
PCA
Principal Component Analysis is a linear dimensionality reduction method that orthogonalizes variables along the directions of maximum variance using singular value decomposition.
LLE
Locally Linear Embedding reconstructs each point as a linear combination of its neighbors and preserves these coefficients in the low-dimensional space.
Nearest Neighbors Graph
Graph connecting each point to its k nearest neighbors, fundamental in UMAP and other manifold methods for approximating local topological structure.
TriMap
Triplets-based Large-Scale Neighbor Embedding uses triplet constraints and negative sampling to preserve local and global structure with less optimization.
Perplexity
Parameter controlling the effective width of the Gaussian kernel in t-SNE and UMAP, balancing between preserving local structure and attention to global variations.
Riemannian Metric
Local geometric structure on a manifold allowing measurement of distances and angles, exploited by UMAP to approximate local differential structure.