Binary Training Policies

The siblings policy is used by default on the local classifier per node, but the remaining ones can be selected with the parameter binary_policy, for example:

• Exclusive
• Less exclusive
• Less inclusive
• Inclusive
• Siblings
• Exclusive siblings

rf = RandomForestClassifier()
classifier = LocalClassifierPerNode(local_classifier=rf, binary_policy="exclusive")

rf = RandomForestClassifier()
classifier = LocalClassifierPerNode(local_classifier=rf, binary_policy="less_exclusive")

rf = RandomForestClassifier()
classifier = LocalClassifierPerNode(local_classifier=rf, binary_policy="less_inclusive")

rf = RandomForestClassifier()
classifier = LocalClassifierPerNode(local_classifier=rf, binary_policy="inclusive")

rf = RandomForestClassifier()
classifier = LocalClassifierPerNode(local_classifier=rf, binary_policy="siblings")

rf = RandomForestClassifier()
classifier = LocalClassifierPerNode(local_classifier=rf, binary_policy="exclusive_siblings")

In the code below, the inclusive policy is selected. However, the code can be easily updated by replacing lines 20-21 with the examples shown in the tabs above.

See also

Mathematical definition on the different policies is given at Training Policies.

Out:

[['Animal' 'Mammal' 'Sheep']
 ['Animal' 'Mammal' 'Sheep']
 ['Animal' 'Mammal' 'Sheep']
 ['Animal' 'Mammal' 'Sheep']]

from sklearn.ensemble import RandomForestClassifier

from hiclass import LocalClassifierPerNode

# Define data
X_train = [[1], [2], [3], [4]]
X_test = [[4], [3], [2], [1]]
Y_train = [
    ["Animal", "Mammal", "Sheep"],
    ["Animal", "Mammal", "Cow"],
    ["Animal", "Reptile", "Snake"],
    ["Animal", "Reptile", "Lizard"],
]

# Use random forest classifiers for every node
# And inclusive policy to select training examples for binary classifiers.
rf = RandomForestClassifier()
classifier = LocalClassifierPerNode(local_classifier=rf, binary_policy="inclusive")

# Train local classifier per node
classifier.fit(X_train, Y_train)

# Predict
predictions = classifier.predict(X_test)
print(predictions)