Scribd Logo
Upload

Welcome to Scribd!

  • 11 views

    Scikit-Learn Integration in The Python Ecosystem

    Uploaded by

    Lenynquiroga

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Scikit-Learn Integration in The Python Ecosystem

    Uploaded by

    Lenynquiroga
    11 views1 page

    Original Title

    sklearn-nips-mloss

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    11 views1 page

    Scikit-Learn Integration in The Python Ecosystem

    Uploaded by

    Lenynquiroga

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 1

    Scikit-Learn: Machine Learning in the Python ecosystem

    Gilles Louppe University of Liège, Belgium


    Gaël Varoquaux Parietal, INRIA Saclay, France

    Scikit-Learn Integration in the Python ecosystem


    The scikit-learn 12 project [4] is an increasingly pop- The library has been designed to tie in with standard
    ular machine learning library written in Python. It open source tools of the scientific Python ecosystem.
    is designed to be simple and efficient, useful to both In particular, scikit-learn leverages NumPy [6] for ef-
    experts and non-experts, and reusable in a variety of ficient storage and manipulation of multi-dimensional
    contexts. The primary aim of the project is to provide arrays, and SciPy [3] for more specialized data struc-
    a compendium of efficient implementations of classic, tures (e.g. sparse matrices) and implementations of
    well-established machine learning algorithms. Among lower-level scientific algorithms. The scikit-learn API
    other things, it includes classical supervised and unsu- is designed to avoid the proliferation of framework
    pervised learning algorithms, tools for model evalua- code: it is limited and non-intrusive. As such it makes
    tion and selection, as well as tools for data preprocess- scikit-learn easy to use and easy to combine with other
    ing and feature engineering. scikit-learn is distributed libraries. Together with IPython [5] for interactive ex-
    under the 3-clause BSD license, encouraging its free ploration and Matplotlib [2] for dynamic data visual-
    use in both commercial and academic settings. ization, NumPy and SciPy constitute a comprehen-
    sive scientific working environment that scikit-learn
    Started in 2007, scikit-learn is developed by an in-
    smoothly complements with a host of machine learning
    ternational team of over a dozen core developers,
    algorithms and data analysis routines.
    mostly researchers from various fields of science. The
    project also benefits from many occasional contribu-
    tors proposing small bugfixes or improvements. De- Demonstrations
    velopment proceeds on GitHub, which greatly facili-
    This presentation will illustrate the use of scikit-learn
    tates this kind of collaboration. Because of the large
    as a component of the larger scientific Python environ-
    number of developers, emphasis is put on keeping the
    ment to solve complex data analysis tasks. Examples
    project maintainable. Code must follow quality guide-
    will include end-to-end workflows based on powerful
    lines, such as style consistency and unit-test coverage.
    and popular algorithms in the library. Among others,
    Documentation and examples are required for all fea-
    we will show how to use out-of-core learning with on-
    tures, and major changes must pass code review by
    the-fly feature extraction to tackle very large natural
    developers not involved in the proposed change.
    language processing tasks, how to exploit an IPython
    All algorithms within scikit-learn are offered through a cluster for distributed cross-validation, or how to build
    simple and elegant API [1] consisting of a well-defined and use random forests to explore biological data.
    set of methods. This API consistency across the pack-
    age makes it very usable in practice: experimenting References
    with different learning algorithm is as simple as sub-
    [1] L. Buitinck et al. API design for machine learning software:
    stituting a class definition. Through composition in- experiences from the scikit-learn project. In ECML/PKDD
    terfaces, the library also offers powerful mechanisms to Workshop: Languages for Data Mining and Machine Learn-
    ing, 2013.
    express a wide variety of learning tasks within a small
    [2] J. D. Hunter. Matplotlib: A 2d graphics environment. CiSE, 9:
    amount of easy-to-read code. Finally, through duck- 90, 2007.
    typing, the consistent API leads to a library that is [3] T. E. Oliphant. Python for scientific computing. CiSE, 9:10,
    2007.
    easily extensible, and allows user-defined estimators to
    [4] F. Pedregosa et al. Scikit-learn: Machine learning in Python.
    be incorporated into the scikit-learn workflow without JMLR, 12:2825–2830, 2011.
    any explicit object inheritance. [5] F. Perez and B. E. Granger. IPython: a system for interactive
    scientific computing. CiSE, 9:21, 2007.
    1
    http://scikit-learn.org [6] S. van der Walt et al. The NumPy array: a structure for efficient
    2
    http://mloss.org/software/view/240 numerical computation. CiSE, 13:22, 2011.

    You might also like