liboml 0.0.2

Library For Online Machine Learning

LIBOML:Library For Online Machine Learning

Here we implement representative online learning algorithms. We split these algorithms into 3 parts:

• Linear classifiers;
• Kernel classifiers;
• Deep classifiers.

The models we have implement:

Name	Binary/Multiclass	Description
OGDBinaryClassifier	binary	Online gradient descent algorithm with logistic loss.
OGDClassifier	multiclass	Online gradient descent algorithm with cross-entropy loss.
PAClassifier	binary	Passive-aggresive algorithm with 3 types of step-size calculation.
Perceptron	binary	Perceptron algorithm.
KernelOGD	binary	Online gradient descent with kernels using logistic loss.
KernelPercetron	binary	Perceptron with kernels.
KernelSVM	binary	Online SVM with kernels.
MLPClassifier	multiclass	Online multi-layer percetron with cross-entropy loss.
ODLClassifier	multiclass	Online deep learning framework.
ODLAEClassifier	multiclass	Online deep learning with Auto-Encoder.

Install

pip install liboml

or

git clone https://github.com/Kaslanarian/liboml
cd liboml
python setup.py install

How to use

Our implementation is based on sklearn, so you can easily use it just like this:

from sklearn.dataset import load_breast_cancer
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split

from liboml.linear import OGDClassifier

X, y = load_breast_cancer(return_X_y=True)
train_X, test_X, train_y, test_y = train_test_split(
    X, 
    y, 
    train_size=0.8,
	random_state=42,
)
stder = StandardScaler().fit(train_X)
train_X, test_X = stder.transform(train_X), stder.transform(test_X)

model = OGDClassifier(init_lr=0.1)
model.fit(train_X, train_y)
acc = model.score(test_X, test_y)
print(acc) # 0.9649122807017544

Future

• More algorithms.
• Support GPU.