Tulia: a comprehensive machine learning project entirely from scratch, utilizing the power of Python and numpy.
By encapsulating both the training and predicting logic within just a couple of classes, complexity is greatly reduced compared to popular frameworks that heavily rely on abstraction. Moreover, the library provided here offers a streamlined approach by maintaining only essential parameters in the model class.
This library uses sklearn API to build the codebase.
from src.linear import LinearRegression
X_train, X_test, y_train, y_test = ...
lr = LinearRegression(n_steps=10_000, learning_rate=1e-4)
lr.fit(X_train, y_train)
y_pred = lr.predict(X_test)
mse = mean_squared_error(y_pred, y_test) # Here mean_squared_error() is a pseudocode.pip install tuliagit clone https://github.com/chuvalniy/Tulia.git
pip install -r requirements.txtEvery machine learning model is provided with unit test that verifies correctness of fit and predict methods.
Execute the following command in your project directory to run the tests.
pytest -vThis demo folder contains jupyter-notebooks that compare scikit-learn and Tulia performance.