def evalFold(arg):
train_index, test_index = arg
SS = StrategyScheduler()
SS.fit(X[train_index], ys[train_index], yNames)
success_count = 0
for i in test_index:
success, score = SSS.eval_schedule('%s#%s' % (XNames[i], SS.predict(X[i])))
if success:
success_count += 1
print 'Current: %i / %i' % (success_count, len(test_index))
return (success_count, len(test_index))
from StrategyScheduler import StrategyScheduler
from MLiP_eval import StrategyScheduleScore
if __name__ == '__main__':
trainFile = 'orig/MLiP_train'
testFile = 'orig/MLiP_train'
mySchedule = 'My_MLiP_train_example_schedule'
SS = StrategyScheduler()
SS.fit_file(trainFile)
X, XNames = StrategyScheduler.read(testFile, False)
N,M = X.shape
# Create schedules
with open(mySchedule,'w') as OS:
for i in range(N):
schedule = SS.predict(X[i])
OS.write('%s#%s\n' % (XNames[i], schedule))
# Evaluate schedule
Eval = StrategyScheduleScore(testFile)
solved, score = Eval.eval_schedules(mySchedule)
print 'Solved: %s %%' % solved
print 'Score: %s %% (%s / %s)' % (round(100*score/Eval.bestScore,4),score,Eval.bestScore)
SS = StrategyScheduler()
SS.fit(X[train_index], ys[train_index], yNames)
success_count = 0
for i in test_index:
success, score = SSS.eval_schedule('%s#%s' % (XNames[i], SS.predict(X[i])))
if success:
success_count += 1
print 'Current: %i / %i' % (success_count, len(test_index))
return (success_count, len(test_index))
if __name__ == '__main__':
trainFile = 'orig/MLiP_train'
SSS = StrategyScheduleScore(trainFile)
X, ys, XNames, yNames = StrategyScheduler.read(trainFile)
N,M = X.shape
kf = skcv.KFold(n=N, n_folds=N-1)
success_count = 0
total_count = 0
pool = Pool()
for success_sub, total_sub in pool.map(evalFold, kf):
success_count += success_sub
total_count += total_sub
print 'Solved: %s %%' % round(100.0 * success_count / total_count, 2)
from StrategyScheduler import StrategyScheduler
from MLiP_eval import StrategyScheduleScore
if __name__ == '__main__':
trainFile = 'orig/MLiP_train'
X, ys, XNames, yNames = StrategyScheduler.read(trainFile)
SS = StrategyScheduler()
SS.fit(X, ys, yNames)
SS.analyze(X, ys)
