TypeError when supplying a boolean X to HuberRegressor fit

[rowan-stein]

User Request

A TypeError is raised when fitting HuberRegressor with boolean predictors. The expectation is that boolean arrays should be accepted and internally converted to float (consistent with LinearRegression).

Specification

• Root cause: HuberRegressor.fit does not coerce boolean inputs to floating-point during validation; downstream numerical routines (L-BFGS-B via SciPy) expect float arrays and error on bool dtype.
• Required change: Validate inputs using scikit-learn’s standard utilities with dtype=FLOAT_DTYPES (via check_X_y(..., dtype=FLOAT_DTYPES, ...)). Handle sample_weight similarly with check_array(..., dtype=FLOAT_DTYPES) and check_consistent_length.
• File/Location: sklearn/linear_model/huber.py, method HuberRegressor.fit(self, X, y, sample_weight=None).
• Sparse handling: Keep current behavior (accept_sparse=['csr']).
• Tests: Add unit tests under sklearn/linear_model/tests/test_huber.py to verify boolean X acceptance and equivalence to float-cast. Include a sample_weight boolean case, consistent with current support.

Steps/Code to Reproduce

import numpy as np
from sklearn.datasets import make_regression
from sklearn.linear_model import HuberRegressor

# Random data
X, y, coef = make_regression(n_samples=200, n_features=2, noise=4.0, coef=True, random_state=0)
X_bool = X > 0
X_bool_as_float = np.asarray(X_bool, dtype=float)

# Works
huber = HuberRegressor().fit(X, y)
# Fails (!)
huber = HuberRegressor().fit(X_bool, y)
# Also works
huber = HuberRegressor().fit(X_bool_as_float, y)

Expected Results

No error is thrown when dtype of X is bool (.fit(X_bool, y) should work). Boolean arrays are expected to be converted to float by HuberRegressor.fit (consistent with LinearRegression).

Actual Results

A TypeError is raised:

TypeError                                 Traceback (most recent call last)
----> 1 huber = HuberRegressor().fit(X_bool, y)

.../sklearn/linear_model/huber.py in fit(self, X, y, sample_weight)
    286             args=(X, y, self.epsilon, self.alpha, sample_weight),
    287             maxiter=self.max_iter, pgtol=self.tol, bounds=bounds,
--> 288             iprint=0)
    289         if dict_['warnflag'] == 2:
    290             raise ValueError("HuberRegressor convergence failed:")

Proposed Fix

• In HuberRegressor.fit, call check_X_y(X, y, dtype=FLOAT_DTYPES, y_numeric=True, accept_sparse=['csr']) to coerce bool inputs to float.
• If sample_weight is provided, validate/cast it using check_array(..., ensure_2d=False, dtype=FLOAT_DTYPES) and verify length consistency with check_consistent_length.