Changes between Version 4 and Version 5 of MatrixFactorization
 Timestamp:
 05/06/11 22:14:53 (3 years ago)
Legend:
 Unmodified
 Added
 Removed
 Modified

MatrixFactorization
v4 v5 1 2 1 = Matrix Factorization Techniques for Data Mining = 3 2 … … 34 33 Reference: (Salakhutdinov, 2008). 35 34 36 === nnma ===37 Nonnegative matrix approximation. Method for dimensionality reduction with respect on the nonnegativity of input data. Multiplicative iterative scheme.[[BR]]38 Reference: (Sra, 2006).39 40 35 === psmf === 41 36 Probabilistic Sparse MF. PSMF allows for varying levels of sensor noise, uncertainty in the hidden prototypes used to explain the data and uncertainty as to the prototypes selected to explain each data vector.[[BR]] … … 43 38 44 39 === bd === 45 Bayesian decomposition. A Bayesian treatment of NMF, based on a normal likelihood and exponential priors, Gibbs samplerto approximate the posterior density.[[BR]]40 Bayesian decomposition. A Bayesian treatment of NMF, based on a normal likelihood and exponential priors, MCMC sampling method (Gibbs) to approximate the posterior density.[[BR]] 46 41 Reference: (Schmidt, 2003). 47 42 48 === bfrm ===49 Bayesian factor regression model. Markov chain Monte Carlo technique.[[BR]]43 === icm === 44 Bayesian model  Iterated conditional modes (ICM) algorithm. The modes of conditional densities have closed form expressions and ICM algorithm has a low computational cost per iteration. [[BR]] 50 45 Reference: (Schmidt, 2003). 51 52 === inmf ===53 Intervalvalued NMF.[[BR]]54 Reference: (Shen, 2010).55 56 === ipmf ===57 Intervalvalued PMF.[[BR]]58 Reference: (Shen, 2010).59 60 46 61 47 == Timeline == … … 82 68 83 69 '''June 18  July 5''' 84 * Implementing family of NMF techniques: sNMF, lNMF, NNMA,PMF.70 * Implementing family of NMF techniques: sNMF, lNMF, PMF. 85 71 86 72 '''July 5  July 15''' … … 93 79 94 80 '''July 15  July 25''' 95 * Implement Bayesian methods. Bayesian decomposition using Gibbs sampler , MCMC techniques such Bayesian factor regression modeling (BD, BFRM).81 * Implement Bayesian methods. Bayesian decomposition using Gibbs sampler (BD). 96 82 * Handling PMF on large, sparse and unbalanced datasets (algorithm for probabilistic sparse matrix factorization (PSMF)). 97 83 98 84 '''July 25  July 31''' 99 * Adapt PMF model to the intervalvalued matrices and implement Intervalvalued PMF (IPMF) and Intervalvalued NMF (INMF).85 * Implement Bayesian model  Iterated conditional modes algorithm (ICM). 100 86 * Improve efficieny of the code, bug removal, exception handling, additional testing. 101 87 … … 112 98 113 99 * Extend Bayesian methods (variational BD, linearly constrained BD). 100 * Adapt PMF model to the intervalvalued matrices and implement Intervalvalued PMF (IPMF) and Intervalvalued NMF (INMF) (Shen, 2010). 101 * Nonnegative matrix approximation. Method for dimensionality reduction with respect on the nonnegativity of input data. Multiplicative iterative scheme (Sra, 2006). 114 102 115 103 