A multi-task version of SConES, which achieves multi-task feature selection coupled with multiple network regularizers using a maximum-flow algorithm.

Please see the following paper for detailed information:

• M. Sugiyama, C.-A. Azencott, D. Grimm, Y. Kawahara, K. M. Borgwardt: Multi-Task Feature Selection on Multiple Networks via Maximum Flows, Proceedings of the SIAM International Conference on Data Mining (SDM 2014), 199-207, 2014 [PDF]

To load files, type in R (without the >, which signifies the prompt):

> source("make.R")
> make()

To run Multi-SConES, type in R:

> mscones(g = g, X = X, Y = Y, lambda = lambda, eta = eta, mu = mu)

• Two R packages igraph and glmnet need to be installed
• g is a graph (in igraph format)
• X is a data matrix (rows: objects, columns: features, each feature corresponds to each vertex in g)
• Y is a matrix of response vectors (rows: objects, columns: tasks)
• lambda, eta, mu are parameters (they should be determined by grid-search with cross-validation)
• output: selected features for each task

> source("make.R")
> make()
> d1 <- generate.data(200, 1, seed = 1)
> d2 <- generate.data(200, 2, seed = 1)
> X <- d1$x; Y <- cbind(d1$y, d2$y)
# simulate two tasks d1$y and d2$y, and d1$x and d2$x are the same
# features from 1 to 44 are causal
> g <- generate.graph()
> res <- mscones(g = g, X = X, Y = Y)
> res
    $`selected features for task 1`
     [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
     [26] 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44

    $`selected features for task 2`
     [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
     [26] 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44

• Author: Mahito Sugiyama
• Affiliation: ISIR, Osaka University
• E-Mail: [email protected]