ClassifierFunction

Status: Stable

documented, exercised by the test suite and/or worked examples, with no known limitations recorded.

Description

ClassifierFunction[method, parameters, featureCount, k]

is the fitted classifier Classify returns. Apply it to a feature vector to get a class, or to a feature vector and "Probabilities" to get one rule per class.

Examples (2)

Every input below was run against the current Mathilda build and its output recorded.

Applications (2)

Classify returns a ClassifierFunction

In[1]:= Head[Classify[{1 -> "low", 2 -> "low", 9 -> "high", 10 -> "high"}]]
Out[1]= ClassifierFunction

Apply it to a new point

In[2]:= c = Classify[{1 -> "low", 2 -> "low", 9 -> "high", 10 -> "high"}]; c[1.5]
Out[2]= "low"

Implementation notes

Definition. ClassifierFunction[method, parameters, featureCount, k] is the fitted classifier object that Classify returns. It is an inert head (no builtin of its own, just a docstring and the Protected attribute); it is produced by Classify and consumed by the prediction machinery in src/ml/predict.c when you apply it.

Representation. A four-argument EXPR_FUNCTION: a method string (e.g. "NearestNeighbors", "LogisticRegression", "NaiveBayes"), a parameters block holding the fitted model (for nearest-neighbours this is the class-label list followed by the labelled training rows), the integer featureCount, and k, the number of classes. Because the object can also be typed by hand, predict.c treats its contents as untrusted and validates them before use.

Usage & limits. Protected. Apply it to a feature vector to get the predicted class; apply it to a feature vector and "Probabilities" to get one class -> p rule per class. It also answers property queries such as "Method" and "FeatureCount". Classify declines a single-class training set (not a classification problem), so a ClassifierFunction always has k >= 2. It is reusable on inputs it was not trained on — that is the whole point of returning a function rather than labels.

Attributes: Protected.

References

• Source: src/ml/classify.c
• Specification index: Mathilda_spec.md
• Tests: tests/test_ml_classify.c

Notes & additional examples

Notes

ClassifierFunction[method, parameters, featureCount, k] is the fitted classifier that Classify returns — a reusable object, not a user-written head. Apply it to a feature vector to get the predicted class, or to a vector and "Probabilities" to get one class -> p rule per class; it also answers "Method" and "FeatureCount".

The four stored parts are the method string (e.g. "NearestNeighbors"), the fitted parameters (for nearest-neighbours, the class-label list followed by the labelled training rows), the input feature count, and k, the number of classes. Classify declines a single-class training set — that is not a classification problem — so a ClassifierFunction always has k >= 2. It is Protected.