BUG SGDRegressor only using the first two samples for validation error calculation #23255
Description
Describe the bug
BaseSGD._make_validation_split returns a 0-1 integer array as validation mask, which is used to calculate the validation score with:
_ValidationScoreCallback(
self,
X[validation_mask],
y[validation_mask],
sample_weight[validation_mask],
classes=classes,here X[validation_mask] is just repeating the first two samples as shown in the following code. A quick fix would be convert validation_mask to a boolean array.
Steps/Code to Reproduce
I added a print statement in _ValidationScoreCallback to print out the shape of validation data
class _ValidationScoreCallback:
"""Callback for early stopping based on validation score"""
def __init__(self, estimator, X_val, y_val, sample_weight_val, classes=None):
self.estimator = clone(estimator)
self.estimator.t_ = 1 # to pass check_is_fitted
if classes is not None:
self.estimator.classes_ = classes
self.X_val = X_val
self.y_val = y_val
self.sample_weight_val = sample_weight_val
print(X_val.shape, y_val.shape)and afterwards running the following code
import numpy as np
from sklearn.linear_model import SGDRegressor
X = np.random.randn(1000, 5)
y = np.random.randn(1000)
sgd = SGDRegressor(early_stopping=True, max_iter=1, validation_fraction=0.1)
sgd.fit(X, y)Expected Results
The size of validation data should be (100,5), (100, ).
Actual Results
prints out (1000, 5), (1000,)
Versions
System:
python: 3.8.8 (default, Apr 13 2021, 12:59:45) [Clang 10.0.0 ]
executable: /Users/mac/Documents/GitHub/.ENV/bin/python
machine: macOS-10.16-x86_64-i386-64bit
Python dependencies:
sklearn: 1.2.dev0
pip: 20.2.3
setuptools: 60.9.3
numpy: 1.21.2
scipy: 1.8.0
Cython: 0.29.28
pandas: 1.3.3
matplotlib: 3.5.1
joblib: 1.1.0
threadpoolctl: 3.0.0
Built with OpenMP: False