GSoC 2018 project ideas

Charles Harris edited this page Mar 7, 2018 · 6 revisions


This is the Google Summer of Code 2018 (GSoC'18) ideas page for SciPy. The SciPy library is one of the core packages that make up the SciPy stack for scientific computing with Python. It provides many user-friendly and efficient numerical routines such as routines for numerical integration and optimization.

This page lists a number of ideas for Google Summer of Code projects for SciPy, plus gives some pointers for potential GSoC students on how to get started with contributing and putting together their application.

Guidelines & requirements

SciPy will participate in GSoC'18 under the umbrella of Python Software Foundation. Please read that page carefully, it contains important advice on the process.

Also read advice on writing a proposal (written with the Mailman project in mind, but generally applicable)

We expect from students that they're at least comfortable with Python (intermediate level). Some projects may also require Cython or C skills. Knowing how to use Git is also important; this can be learned before the official start of GSoC if needed though.

Getting started on SciPy

Potential candidates should take a look at the guidelines on how to contribute to Scipy. Making a small enhancement/bugfix/documentation fix/etc (does not need to be related to your proposal) to Scipy before applying for the GSoC is a hard requirement; it can help you get some idea how things will work during GSoC.

Start on your proposal early, post a draft to the mailing list and iterate based on the feedback you receive. This will not only improve the quality of your proposal, but also help you find a suitable mentor.

SciPy Project ideas

Enhance the Randomized Numerical Linear Algebra functionality

In SciPy 1.0 a first randomized linear algebra method was added, namely CountMin Sketch (clarkson_woodruff_transformation) for reducing the dimensionality of a vector space to an embedded space.

Other widely used methods for subspace embedding (like the Johnson-Lindenstrauss) can be added, as well as algorithms that build on top of those subspace embedding methods for least squares regression or low rank approximation.

  • Required knowledge: Python, linear algebra
  • Difficulty level: medium
  • Potential mentors: Ralf Gommers

Rotation formalism in 3 dimensions

General info

  • Required knowledge and skills: Python, understanding how to vectorize operations in numpy, ability to read and grasp papers/books on the subject
  • Difficult level: easy to medium
  • Potential mentors: Eric Larson, Nikolay Mayorov


There are several ways to represent a rotation in 3-dimensional space: Euler angles, direction cosine (or rotation) matrices (DCM), quaternions, rotation vectors and some others. Different representations are preferable in different scenarios and there exists a correspondence between all of them. You can find a good overview in [1] and a more comprehensive treatment in [2]. Note that there are many different conventions in this field, one of which should be chosen and described properly. Also note that "rotation" and "orientation" can be used interchangeably: an orientation between two frames is determined by a rotation required to superimpose one frame onto another.

The aim of this project is to create a module which will allow to conveniently describe, apply and compose rotations. The proposed module name is scipy.spatial.rotation or if are looking for wider perspectives scipy.spatial.transform. The content of this module is preliminarily seen as follows:

  1. The main class of the module may be called Rotation.
    1. It should represent an arbitrary rotation or several rotations (i.e. we should aim for vectorized approach for efficiency).
    2. It should be possible to create Rotation from any sensible representation and in vectorized fashion. The most widely used of them are listed above, more exotic parametrizations can also be considered.
    3. Internally Rotation should use either DCMs or quaternions to represent rotations. Both of these choices are good with some minor pros and cons.
    4. Rotation should be able to return any required representation, like for example we would like to know which Euler angles describe our rotation, etc. These representations can be returned as numpy arrays.
    5. Rotation should have methods to rotate or project a vector or a set of vectors. See [3] to get an idea about the difference of this views. Overall it should be intuitive and clear for the user what he is doing with which operation (i.e. it should be explained sufficiently in the documentation). Also here we need to consider how vectorized operations are performed, i.e. one rotation on many vectors, many rotations on one vector and many rotations on many vectors (most likely we need 1-to-1 correspondence in this case).
    6. Rotation should have a method (likely __mul__) to compute a representation of a consecutive application of two rotation, like C = A * B. If DCMs are used internally this composition comes down simply to matrix multiplication.
    7. The main difficulty in implementing this class is to implement conversions between an arbitrary representation to/from the internal representation. There are some peculiar points.
  2. Apart from Rotation class we can include some useful algorithms regarding rotations. Some that I'm aware of:
    1. Interpolation between orientations using SLERP algorithm [4].
    2. Finding a rotation which gives the best match between the set of vectors measured in two frames. Note that some methods produce general orthogonal transformations, i.e., allow reflections [5].
    3. "Cubic spline" interpolation for the set of orientations [6].
    4. Optimal "averaging" of orientations [7].
    5. Orthogonalization of a 3x3 matrix, one of simpler approaches is described in [8].
    6. Uniform random sampling of rotations. One way to achieve this is to sample quaternions as unit vectors distributed uniformly on a unit sphere in 4D.


  2. "A Survey of Attitude Representation" by Malcolm D. Shuster
  7. "Quaternion Averaging" by F. Landis Markley et al.
  8. "Unit Quaternion from Rotation Matrix" by F. Landis Markley

Your own idea

The above projects are just suggestions --- it is also very good to suggest a project idea of your own if you have something in mind that you want to do. Ask people on the mailing list for suggestions in this case.

And more...

A set of additional new feature ideas can be found on the Scipy Roadmap.

You can’t perform that action at this time.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session.
Press h to open a hovercard with more details.