lindahua / pli-toolbox Goto Github PK

• a common function symeig
• PCA: prin_comp and prin_comp_cov
• Fisher LDA: linear_discrim
• CCA (canonical correlation analysis): canon_corr

• HMM inference
• HMM training
• Viterbi decoding

Add more functions to the metrics module:

• pw_euclidean
• pw_cityblock
• pw_chebyshev
• pw_hamming
• pw_cosine
• pw_qdist
• pw_mahdist
• pw_minkowski

• graphcut solver
• loopy belief propagation algorithm

Refer to a NIPS 2011 paper (by Bergstra et al.):

Algorithms for Hyper-Parameter Search

• pli_addpath
• pli_build

• aggregation
• counting
• grouping
• binning

This includes

• a PPCA class to support single probabilistic pca computation
• a PPCA model model
• mixture of PPCAs

-- plot hard-labeled points using colors
-- plot soft-labeled points using mix of colors
-- plot lines given theta and b

Several commonly used sampling function

• find_bin: find the bin label for a number
• sample_wor: Simple sampling without replacement
• ddsample: sampling from discrete distribution

Vector quantization clustering methods

• seeding: rand_seed, kmpp_seed
• kmeans
• kmedoid

• Mahalanobis distances
• logpdf evaluation
• samping
• maximum likelihood estimation

• linear SVM
• kernel evaluation
• kernel SVM

• covariance and entropy evaluation
• logpdf evaluation
• sampling
• maximum likelihood estimation

• Fixed training and testing set
• n-fold cross validation
• leave-one-out cross validation

• a general mixture model framework
• Gaussian mixture model
• a demo of GMM

• Gauss functions using canonical parameters
• functions to generate weight matrices and Laplacian matrix
• GMRF-based smoothing