Skip to content

Support VFDT trees #9

Description

@anjsimmo

River_return_condition tests for EFDTNumericBinaryBranch. This needs to be generalised to also support VFDT (HoeffdingTreeClassifier) which uses NumericBinaryBranch (which EFDTNumericBinaryBranch inherits from)

def river_return_condition(node):
    if isinstance(node, river.tree.nodes.efdtc_nodes.EFDTNumericBinaryBranch):
        return Condition(f"attr_{node.feature}", Operator.LE, node.threshold)
    elif isinstance(node, tuple):  # Multinomial
        feature = node[0].feature
        threshold = node[0]._r_mapping[node[1]]
        return Condition(f"attr_{feature}", Operator.EQ, threshold)
    else:
        raise ValueError(node)

This results in the following exception when run on a HoeffdingTreeClassifier rather than a ExtremelyFastDecisionTreeClassifier

Traceback (most recent call last):
  File "/home/anj/src/paper-decision-trees/tree_diff2/experiment2.py", line 260, in eval_efdt
    batch1_rules = Ruleset(river_extract_rules(model_batch._root,river_children, river_is_leaf))
  File "/home/anj/src/paper-decision-trees/tree_diff2/tree_diff/tree_ruleset_conversion.py", line 137, in river_extract_rules
    return [river_create_rule(p) for p in walk_tree(tree, children, is_leaf)]
  File "/home/anj/src/paper-decision-trees/tree_diff2/tree_diff/tree_ruleset_conversion.py", line 137, in <listcomp>
    return [river_create_rule(p) for p in walk_tree(tree, children, is_leaf)]
  File "/home/anj/src/paper-decision-trees/tree_diff2/tree_diff/tree_ruleset_conversion.py", line 133, in river_create_rule
    return Rule(conditions=river_create_conditions(path[0:-1]), label=f"{label}")
  File "/home/anj/src/paper-decision-trees/tree_diff2/tree_diff/tree_ruleset_conversion.py", line 123, in river_create_conditions
    return [river_return_condition(c) for c in path_conds]
  File "/home/anj/src/paper-decision-trees/tree_diff2/tree_diff/tree_ruleset_conversion.py", line 123, in <listcomp>
    return [river_return_condition(c) for c in path_conds]
  File "/home/anj/src/paper-decision-trees/tree_diff2/tree_diff/tree_ruleset_conversion.py", line 119, in river_return_condition
    raise ValueError(node)
ValueError: NumericBinaryBranch

See the following notebook for a demonstration of the issue: https://github.com/a2i2/tree_diff/blob/re-evaluation/notebooks/Similarity%20Score%20Issues%20-%20VFDT.ipynb

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions