Factors, properties, or structures can correlate with response values {regression algorithms}.
regression
Regression analysis finds property and structure relationships. Multiple linear regression measures linear-component dependence on properties and finds descriptor coefficients. Non-linear regression is a parametric method that finds descriptor coefficients. Ridge regression is another regression method.
correlation
Factors can correlate, with correlation coefficients. Variance-covariance matrices {correlation matrix, vision} are complete, symmetric, square matrices that use property values and structure values, which can scale to normalize data. Partial least-squares can simplify variance-covariance matrix {matrix diagonalization, vision} {matrix bidiagonalization method, regression}.
Spearman rank correlation coefficient can measure molecular similarity.
least-squares
Ordinary least-squares {classical least-squares, vision} {least-squares regression, vision} {linear least-squares regression, vision} {multiple least-squares regression, vision} {multivariate least-squares regression, vision} can find descriptor coefficients by minimizing distances between values and regression line. Inverse least-squares inverts fitting method.
adaptive least-squares
Adaptive least-squares modifies ordinary least-squares by weighting or classes.
adaptive least-squares: fuzzy
Features can be in different classes with different weights.
partial least-squares
PLS uses least-squares to find independent variables and dependencies among variables. PLS maximizes latent-variable and observable covariation. PLS diagonalizes variance-covariance matrix. Multi-block PLS uses groups. Kernel algorithm is about covariation.
partial least-squares: Comparative Molecular Field Analysis
Partial least-squares methods (CoMFA) can analyze grids around sites and find grid-point interactions, to make sampled-point descriptors.
partial least-squares: Generating Optimal Linear PLS Estimations
GOLPE uses PLS and D-optimal design to select variables and then cross-validates.
partial least-squares: SAMPLS algorithm
Partial least-squares and trend vector analysis can work together.
non-least-squares
Non-least-squares methods can detect non-linear relationships.
3-Computer Science-Systems-Computer Vision-Algorithms
Outline of Knowledge Database Home Page
Description of Outline of Knowledge Database
Date Modified: 2022.0225