def testAdaptive(dataPackage, time, target_acc, kfold):
from copy import deepcopy
from AUREA.adaptive.Adaptive import Adaptive
#partition the samples
lq = adaptiveGetLearnerQueue(dataPackage)
adaptive = Adaptive(lq)
return adaptive.crossValidate(target_acc,time, kfold)
def trainAdaptive(self, target_accuracy, maxTime ):
self.queue.put(('statusbarset',"Configuring adaptive training"))
acc = float(target_accuracy)
mtime = int(maxTime)
maxTime = mtime
target_accuracy = acc
#build learner queue
self._adaptiveSetup()
#create adaptive object
adaptive = Adaptive(self.learnerqueue, app_status_bar = self.queue)
top_acc, top_settings, top_learner = adaptive.getLearner(target_accuracy, maxTime)
#store adaptive results (really should be in adaptive)
self.adaptive_history = adaptive.getHistory()
self.adaptive_history.reverse()
self.adaptive = top_learner
self.adaptive_settings = top_settings
self.adaptive_top_mcc = top_acc
self.adaptive_setting_string = adaptive.getSettingString(top_settings)
if self.adaptive is not None:
row_key = top_settings['data_type']
self.queue.put(('statusbarset',"Training Complete, Checking Accuracy"))
self.adaptive_acc = self._getLearnerAccuracy(self.adaptive, row_key)
self.queue.put(('statusbarset',"Accuracy Check Complete"))
else:
#none of the algorithms ran, maybe timeout is to low
self.queue.put(('statusbarset',"Adaptive failed to run. Is the timeout too low?"))
def crossValidateAdaptive(self, target_acc, maxtime):
self._adaptiveSetup()
#create adaptive object
adaptive = Adaptive(self.learnerqueue, app_status_bar = self.queue)
#using accuracy, because we are reporting accuracy - acc is not used to choose learner
adaptive.crossValidate(target_acc, maxtime)
self.adaptive_tt = adaptive.truth_table[:]
self.queue.put(('statusbarset',"Adaptive had an Accuracy of " + str(self._acc(self.adaptive_tt))[:4]))
