CapyMOA (streaming / online-learning) bridge for TuiML.
Importing this package registers the wrapped CapyMOA incremental learners into the TuiML hub under capymoa. keys, so they never collide with the native TuiML algorithms in algorithms — "NaiveBayes" is a CapyMOA-only key, but "AdaptiveRandomForest" would collide without the namespace.
The wrapper modules mirror TuiML's own taxonomy — bayesian, trees, ensemble — so a wrapped learner sits in the same place as its native counterpart:
bayesian — incremental Naive Bayes.
trees — Hoeffding / EFDT trees.
ensemble — online bagging, adaptive random forests.
CapyMOA is an optional dependency, and it needs a Java runtime (11+) on PATH as well, because MOA runs on the JVM:
java -version # must succeed first pip install 'tuiml[capymoa]' # or: uv pip install 'tuiml[capymoa]'
If CapyMOA is not installed, importing this package still succeeds (the wrapper classes are defined), but instantiating a wrapper raises a clear ImportError pointing at pip install 'tuiml[capymoa]'.
Streaming learners update one instance at a time. Use partial_fit to feed
successive batches; it returns self, so calls chain. fit is also
available and simply streams the whole array in one pass:
>>> from tuiml.capymoa import HoeffdingTree, NaiveBayes
>>> from tuiml.datasets.generators import Agrawal
>>> data = Agrawal(n_samples=500, function=1, random_state=42).generate()
>>> # Incremental training across multiple batches
>>> model = HoeffdingTree(grace_period=100).partial_fit(data.X[:300], data.y[:300])
>>> model = model.partial_fit(data.X[300:], data.y[300:])
>>> model.predict(data.X[:3]).shape
(3,)>>> tree = NaiveBayes().partial_fit(data.X[:300], data.y[:300])
>>> tree = tree.partial_fit(data.X[300:], data.y[300:])
>>> tree.predict(data.X[:3]).shape
(3,)They are also addressable by hub key, as "capymoa.<ClassName>", which is
the form train and an agent over MCP both use:
>>> import tuiml
>>> import tuiml.capymoa # registers the capymoa.* keys
>>> model = tuiml.train({
... "model": {"name": "capymoa.HoeffdingTree"},
... "data": {"source": "iris", "target": "class"},
... "evaluation": {"test_size": 0.3},
... })
>>> "accuracy_score" in model.metrics_
True