def competitionRunner(X, Y, schedules, m, n, tFactor, measureStep):
# CONSTANTS
SCHEDULES = schedules
N = n
M = m
T = tFactor * N
MEASURE_STEP = measureStep
RUN_COUNT = 4
THREAD_COUNT = 4
# SEEDS
samplerSeeds = [rng.randint(0,2**32-1) for i in range(RUN_COUNT)]
runnerSeeds = [rng.randint(0,2**32-1) for i in range(RUN_COUNT)]
# INPUT VARIABLES
patterns = X
classes = Y
# SAMPLERS & RUNNERS
samplers = [mcaa.sampler(N, samplerSeeds[i]) for i in range(RUN_COUNT)]
runners = [mcaa.CompetitionRunner(T, MEASURE_STEP, SCHEDULES[i], samplers[i], patterns, classes, runnerSeeds[i]) \
for i in range(RUN_COUNT)]
multiRunner = mcaa.multiCompetitionRunner(THREAD_COUNT, [runner.getPointer() for runner in runners])
# RUN
multiRunner.runAll()
# OUTPUTS
allEMeasures = multiRunner.getEMeasures() / M
allWeights = [sampler.getSamples() for sampler in samplers]
return allEMeasures, allWeights
def linearRunner(offset, slope, alpha, n, tFactor, measureStep, runCount,
threadCount):
# CONSTANTS
OFFSET = offset
SLOPE = slope
ALPHA = alpha
N = n
M = int(ALPHA * N)
T = tFactor * N
MEASURE_STEP = measureStep
RUN_COUNT = runCount
THREAD_COUNT = threadCount
# SEEDS
weightsSeed = rng.randint(0, 2**32 - 1)
patternsSeed = rng.randint(0, 2**32 - 1)
samplerSeeds = [rng.randint(0, 2**32 - 1) for i in range(RUN_COUNT)]
runnerSeeds = [rng.randint(0, 2**32 - 1) for i in range(RUN_COUNT)]
# INPUT VARIABLES
weights = mcaa.buildRandomWeights(N, weightsSeed)
patterns = mcaa.buildRandomPatterns(N, M, patternsSeed)
classes = mcaa.buildClasses(weights, patterns)
# SAMPLERS & RUNNERS
samplers = [mcaa.sampler(N, samplerSeeds[i]) for i in range(RUN_COUNT)]
schedules = [
mcaa.continuousLinearSchedule(T, OFFSET, SLOPE)
for i in range(RUN_COUNT)
]
runners = [
mcaa.MCMCRunner(T, MEASURE_STEP, schedules[i], samplers[i], weights,
patterns, classes, runnerSeeds[i])
for i in range(RUN_COUNT)
]
multiRunner = mcaa.multiMCMCRunner(
THREAD_COUNT, [runner.getPointer() for runner in runners])
# RUN
multiRunner.runAll()
# OUTPUTS
allEMeasures = multiRunner.getEMeasures() / M
allOverlapMeasures = multiRunner.getOverlapMeasures()
return allEMeasures, allOverlapMeasures