def crossval_baseline(folds):
# This is essentially a "one-hit-all" evaluation of policy 4
results = {}
for fold_name, data in folds:
predictions = {}
truths = {}
for datum in data:
prediction = classifier.policy(datum, 6)
truth = datum.label
key = (datum.evt, datum.ctxGrounded)
if key in predictions:
# If the prediction if false, but the value is true,
# lets override it. Otherwise the result is the same
if predictions[key] == 0:
predictions[key] = prediction
if truths[key] == 0:
truths[key] = truth
else:
# Store a result anyway
predictions[key] = prediction
truths[key] = truth
keys = predictions.keys()
result = ClassificationResults(fold_name, [truths[k] for k in keys], [predictions[k] for k in keys], [k for k in keys])
results[fold_name] = result
return results
def crossval_baseline(folds):
# This is essentially a "one-hit-all" evaluation of policy 4
results = {}
for fold_name, data in folds:
predictions = {}
truths = {}
for datum in data:
prediction = classifier.policy(datum, 6)
truth = datum.label
key = (datum.evt, datum.ctxGrounded)
if key in predictions:
# If the prediction if false, but the value is true,
# lets override it. Otherwise the result is the same
if predictions[key] == 0:
predictions[key] = prediction
if truths[key] == 0:
truths[key] = truth
else:
# Store a result anyway
predictions[key] = prediction
truths[key] = truth
keys = predictions.keys()
result = ClassificationResults(fold_name, [truths[k] for k in keys],
[predictions[k] for k in keys],
[k for k in keys])
results[fold_name] = result
return results
def policy(datum):
return classifier.policy(datum, 3)
def policy(datum):
return classifier.policy(datum, 3)
