Intel® Extension for Scikit-learn Ridge Regression for Higgs dataset

from timeit import default_timer as timer
from sklearn import metrics
from sklearn.model_selection import train_test_split
import warnings
from sklearn.datasets import fetch_openml
from sklearn.preprocessing import LabelEncoder
from IPython.display import HTML

warnings.filterwarnings("ignore")

Download the data

dataset = fetch_openml(data_id=45570, as_frame=True)

Preprocessing

Let’s encode categorical features with LabelEncoder

# Access the data and target
x = dataset.data
y = dataset.target

for col in ['jet 3 b-tag', 'jet 1 b-tag', 'jet 4 b-tag', 'jet 2 b-tag']:
    le = LabelEncoder().fit(x[col])
    x[col] = le.transform(x[col])

x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.1, random_state=0)
x_train.shape, x_test.shape, y_train.shape, y_test.shape

((9900000, 28), (1100000, 28), (9900000,), (1100000,))

from sklearn.preprocessing import MinMaxScaler, StandardScaler

scaler_x = MinMaxScaler()
scaler_y = StandardScaler()

y_train = y_train.to_numpy().reshape(-1, 1)
y_test = y_test.to_numpy().reshape(-1, 1)

scaler_x.fit(x_train)
x_train = scaler_x.transform(x_train)
x_test = scaler_x.transform(x_test)

scaler_y.fit(y_train)
y_train = scaler_y.transform(y_train).ravel()
y_test = scaler_y.transform(y_test).ravel()

Patch original Scikit-learn with Intel® Extension for Scikit-learn

Intel® Extension for Scikit-learn (previously known as daal4py) contains drop-in replacement functionality for the stock Scikit-learn package. You can take advantage of the performance optimizations of Intel® Extension for Scikit-learn by adding just two lines of code before the usual Scikit-learn imports:

from sklearnex import patch_sklearn

patch_sklearn()

Intel(R) Extension for Scikit-learn* enabled (https://github.com/uxlfoundation/scikit-learn-intelex)

from sklearn.linear_model import Ridge

params = {
    "alpha": 0.3,
    "fit_intercept": False,
    "random_state": 0,
    "copy_X": False,
}
start = timer()
model = Ridge(random_state=0).fit(x_train, y_train)
train_patched = timer() - start
f"Intel® extension for Scikit-learn time: {train_patched:.2f} s"

'Intel® extension for Scikit-learn time: 0.46 s'

y_predict = model.predict(x_test)
mse_metric_opt = metrics.mean_squared_error(y_test, y_predict)
f"Patched Scikit-learn MSE: {mse_metric_opt}"

'Patched Scikit-learn MSE: 0.9023910206230603'

Train the same algorithm with original Scikit-learn

In order to cancel optimizations, we use unpatch_sklearn and reimport the class Ridge

from sklearnex import unpatch_sklearn

unpatch_sklearn()

from sklearn.linear_model import Ridge

start = timer()
model = Ridge(random_state=0).fit(x_train, y_train)
train_unpatched = timer() - start
f"Original Scikit-learn time: {train_unpatched:.2f} s"

'Original Scikit-learn time: 4.45 s'

y_predict = model.predict(x_test)
mse_metric_original = metrics.mean_squared_error(y_test, y_predict)
f"Original Scikit-learn MSE: {mse_metric_original}"

'Original Scikit-learn MSE: 0.9023910206230603'

HTML(
    f"<h3>Compare MSE metric of patched Scikit-learn and original</h3>"
    f"MSE metric of patched Scikit-learn: {mse_metric_opt} <br>"
    f"MSE metric of unpatched Scikit-learn: {mse_metric_original} <br>"
    f"Metrics ratio: {mse_metric_opt/mse_metric_original} <br>"
    f"<h3>With Scikit-learn-intelex patching you can:</h3>"
    f"<ul>"
    f"<li>Use your Scikit-learn code for training and prediction with minimal changes (a couple of lines of code);</li>"
    f"<li>Get comparable model quality</li>"
    f"<li>Get a <strong>{(train_unpatched/train_patched):.1f}x</strong> speedup.</li>"
    f"</ul>"
)

Compare MSE metric of patched Scikit-learn and original

MSE metric of patched Scikit-learn: 0.9023910206230603
MSE metric of unpatched Scikit-learn: 0.9023910206230603
Metrics ratio: 1.0

With Scikit-learn-intelex patching you can:

• Use your Scikit-learn code for training and prediction with minimal changes (a couple of lines of code);
• Get comparable model quality
• Get a 9.7x speedup.