jnmf.py

from lib.ToyMatrixClass import *
from lib.StandardJnmfClass import *
from lib.IntegrativeJnmfClass import *
import lib.datasets as data
# Mutable Variables
k_list = [3]
lamb = [5]
save = 0
niter = 4960
super_niter = 1
thresh = 1.5
data_set = [
            "simulated",
            "small"
            ]

# Input
x={}
tm = ToyMatrix('m')
x[data_set[0]] = tm.x
x[data_set[1]] = data.small()

data_set = [
            # "simulated",
            "small"
            ]

# Program
print("Super niter: %i Niter: %i" % (super_niter, niter))
k_list.sort()
for d in data_set:
    print("\tDataset: %s" % d)

    # sm = {}
    #
    # print("\t\tStandard NMF")
    # for k in k_list:
    #     sm[k] = StandardNmfClass(x[d], k, niter=niter, super_niter=super_niter, thresh=thresh)
    #
    #     sm[k].super_wrapper(verbose=0)
    #
    #     print("\t\t\tSNMF K = %i Error = %f" % (k, sm[k].error))
    #     heatmap(sm[k].max_class_cm, "SNMF-h-k=%i" % k, folder=d, save=save)
    #     heatmap(sm[k].cmw, "SNMF-w-k=%i" % k, folder=d, save=save)

    im = {}
    i=1
    print("\tIntegrative NMF")
    for l in lamb:
        print("\t\tLambda(L)%i = %f" % (i, l))
        i+=1
        for k in k_list:
            im[k] = IntegrativeNmfClass(x[d], k, niter=niter, super_niter=super_niter, lamb=l, thresh=thresh)

            im[k].super_wrapper(verbose=0)

            print("\t\t\tINMF K = %i Error = %f" % (k, im[k].error))
            heatmap(im[k].max_class_cm, "L%i-INMF-h-k=%i" % (i, k), folder=d, save=save)
            heatmap(im[k].cmw, "L%i-INMF-w-k=%i" % (i, k), folder=d, save=save)