'Time Delay': CRPtimeDelays,
'dA Learning Rate': learningRates,
'dA Batch Size': batchSizes,
'dA Num Hidden Units': numHiddens,
'dA Corruption Level': corruptionLevels}
numRuns = 0
deleteNCDPickleFiles()
existingNCDs = None
processPool = Pool(numProcesses)
iteration = 0
stopRunningAt = datetime(2015, 8, 18, 10)
opt = Optimiser(settingsDict,
oldResultsDataFrame = None,
resultsColumn = 'Mean Average Precision',
noImprovementStoppingRounds = None)
currentDateTime = datetime.now()
while currentDateTime < stopRunningAt:
nextSettings = True
iteration += 1
while nextSettings is not None and currentDateTime < stopRunningAt:
nextSettings = opt.getNextSettings()
if nextSettings is not None:
for setting in nextSettings:
# Create CRPs and NCDs
# load weights if this is for a neural net run
'FENS Quantisation Weight 3': FENSquantisationWeights3,
'FENS Quantisation Weight 4': FENSquantisationWeights4
}
if NNtype is not None:
settingsDict['dA Num Hidden Units'] = numHiddens
settingsDict['dA Num Visible Units'] = numVisibles
settingsDict['dA Corruption Level'] = corruptionLevels
# Initialise
numRuns = 0
iteration = 2
processPool = Pool(numProcesses)
opt = Optimiser(settingsDict,
oldResultsDataFrame=None,
resultsColumn='Mean Average Precision',
noImprovementStoppingRounds=None,
floatRounding=4)
featureFileDict = None
FENSfeatureFileDict = None
currentDateTime = datetime.now()
# Load base features
piecesPath = FFP.getRootPath(baseFeatureName)
pieceIds = getFolderNames(piecesPath,
contains='mazurka',
orderAlphabetically=True)[:numFolders]
print 'Loading feature file dict...'
featureFileDict = FFP.loadFeatureFileDictAllFolders(piecesPath, pieceIds,
baseFeatureName,
parser.add_argument('-o', '--output', default=None)
args = parser.parse_args()
def id_generator(size=6, chars=string.ascii_uppercase + string.digits):
return ''.join(random.choice(chars) for x in range(size))
new_tmpdir = tempfile.mkdtemp(prefix='tmpwrap_mgp_', dir=args.tmpdir)
c = Collection(input_dir=args.input_dir,
compression=args.compression,
file_format=args.format,
datatype=args.datatype,
tmpdir=new_tmpdir)
o = Optimiser(args.nclusters, c)
o.optimise(max_iter=500,
nreassign=args.nreassign,
sample_size=args.sample_size)
output_name = args.output or 'output_' + id_generator(6)
output_fh = open(output_name, 'w+')
headings = ['Iteration', 'CPU Time', 'Likelihood', 'Partition']
output = [[i] + x for i, x in enumerate(o.Scorer.history)]
writer = csv.writer(output_fh, delimiter='\t', quoting=csv.QUOTE_NONE)
writer.writerow(headings)
writer.writerows(output)
parser.add_argument('-c', '--compression', default=None)
parser.add_argument('-t', '--tmpdir', default='/tmp/')
# Collect all args for optimse and parse them later?
parser.add_argument('-r', '--nreassign', default=10, type=int)
parser.add_argument('-s', '--sample_size', default=10, type=int)
parser.add_argument('-o', '--output', default=None)
args = parser.parse_args()
def id_generator(size=6, chars=string.ascii_uppercase + string.digits):
return ''.join(random.choice(chars) for x in range(size))
new_tmpdir = tempfile.mkdtemp(prefix='tmpwrap_mgp_', dir=args.tmpdir)
c = Collection(input_dir=args.input_dir,
compression=args.compression, file_format=args.format, datatype=args.datatype,
tmpdir=new_tmpdir)
o = Optimiser(args.nclusters, c)
o.optimise(max_iter=500, nreassign=args.nreassign, sample_size=args.sample_size)
output_name = args.output or 'output_' + id_generator(6)
output_fh = open(output_name, 'w+')
headings = ['Iteration', 'CPU Time', 'Likelihood', 'Partition']
output = [[i] + x for i, x in enumerate(o.Scorer.history)]
writer = csv.writer(output_fh, delimiter='\t', quoting=csv.QUOTE_NONE)
writer.writerow(headings)
writer.writerows(output)
#!/usr/bin/env python3
from utils import read_matrix, read_data
from optimiser import Optimiser
# Read the data, one number (float) per line, so line 0 has the data for vertex 0 and so-on
data = read_data("data.txt")
# Read an n x n adjacency matrix, each row on a new line and separated by commas
W = read_matrix("adj_matrix.txt")
# Create an optimser object
opt = Optimiser(W, data)
# Call the optimiser with the selected parameters, returning the optimal graph
optimal_graph = opt.iterative_opt(remove=True, add=True, remove_first=False)