【笔记】GraphLab机器学习平台介绍

一个机器学习平台，主要是图模型方面的计算。

第一个Demo就是计算PageRank 。

还提供包括求解线性方程组，协同过滤、聚类等等操作。

GraphLab linear solvers library

This application solves a linear system of equations using iterative solvers: Jacobi, Gaussian Belief Propagation (GaBP), Conjugate gradient, inverse of sparse symmetric matrix via GaBP, Shotgun LASSO solver and Shotgun sparse logistic regression solver.

GraphLab collaborative filtering library

This library implements multiple algorithms for factorizing a 3D tensor or a 2D matrix into lower rank matrices. Implemented algorithms are: PMF (probabalistic matrix factorization), BPTF (Bayesian probablistic tensor factorization), ALS (alternating least squares), WALS (weighted alternating least squares), SGD (stochastic gradient descent), SVD (Lanczos algorithm), NMF (non-negative matrix factorization) and Koren's SVD++ algorithm.

GraphLab clustering library

This library implements multiple clustering methods like K-Means, Fuzzy-Kmeans, Kmeans++, LDA (Latent Dirichlet Allocation), K-Core decomposition.

Non-parametric belief propagation

Code for running Non-parametric belief propagation algorithm, for computing inference using Gaussian mixture model.