'iterations' : 3000,
'init_FG' : 'kmeans',
'init_S' : 'exponential',
'expo_prior' : 0.1
}
K_range = [2,4,6,8,10]
L_range = [2,4,6,8,10]
P = 5
no_folds = 5
output_file = "./results.txt"
files_nested_performances = ["./fold_%s.txt" % fold for fold in range(1,no_folds+1)]
# Construct the parameter search
parameter_search = [{'K':K,'L':L} for (K,L) in itertools.product(K_range,L_range)]
# Load in the Sanger dataset
(_,X_min,M,_,_,_,_) = load_Sanger(standardised=standardised)
# Run the cross-validation framework
random.seed(42)
numpy.random.seed(9000)
nested_crossval = MatrixCrossValidation(
method=NMTF,
X=X_min,
M=M,
K=no_folds,
parameter_search=parameter_search,
train_config=train_config,
file_performance=output_file
)
nested_crossval.run()
output_file = "./results_TEST.txt"
alpha, beta = 1., 1.
lambdaF = 1. / 10.
lambdaS = 1. / 10.
lambdaG = 1. / 10.
priors = {
'alpha': alpha,
'beta': beta,
'lambdaF': lambdaF,
'lambdaS': lambdaS,
'lambdaG': lambdaG
}
# Load in the Sanger dataset
(_, X_min, M, _, _, _, _) = load_Sanger(standardised=standardised)
# Run the cross-validation framework
random.seed(42)
numpy.random.seed(9000)
nested_crossval = GreedySearchCrossValidation(classifier=bnmtf_vb_optimised,
R=X_min,
M=M,
values_K=K_range,
values_L=L_range,
folds=no_folds,
priors=priors,
init_S=init_S,
init_FG=init_FG,
iterations=iterations,
restarts=restarts,
standardised = False #standardised Sanger or unstandardised
repeats = 10
iterations = 500
init_UV = 'random'
I, J, K = 622,139,25
alpha, beta = 1., 1. #1., 1.
lambdaU = numpy.ones((I,K))/10.
lambdaV = numpy.ones((J,K))/10.
priors = { 'alpha':alpha, 'beta':beta, 'lambdaU':lambdaU, 'lambdaV':lambdaV }
# Load in data
(_,R,M,_,_,_,_) = load_Sanger(standardised=standardised)
# Run the VB algorithm, <repeats> times
times_repeats = []
performances_repeats = []
for i in range(0,repeats):
# Set all the seeds
numpy.random.seed(0)
# Run the classifier
BNMF = bnmf_gibbs_optimised(R,M,K,priors)
BNMF.initialise(init_UV)
BNMF.run(iterations)
# Extract the performances and timestamps across all iterations
import numpy, random, scipy, matplotlib.pyplot as plt
##########
standardised = False #standardised Sanger or unstandardised
repeats = 10
iterations = 1000
I, J, K = 622, 139, 25
init_UV = 'exponential'
expo_prior = 1 / 10.
# Load in data
(_, R, M, _, _, _, _) = load_Sanger(standardised=standardised)
# Run the VB algorithm, <repeats> times
times_repeats = []
performances_repeats = []
for i in range(0, repeats):
# Set all the seeds
numpy.random.seed(0)
# Run the classifier
nmf = NMF(R, M, K)
nmf.initialise(init_UV, expo_prior)
nmf.run(iterations)
# Extract the performances and timestamps across all iterations
times_repeats.append(nmf.all_times)