Create Testing Ideas

[Wattsy2020]

Coming up with ideas to test numpy's implementation of mathematical functions. Given numpy implements most of practical mathematics this is simply "come up with an interesting mathematical relation". This can latter be implemented in python code as automated tests.

Some examples of the metamorphic relations required:

• sin(x) = sin(pi - x)
• ab = ba
• the length of a vector x = sqrt(x^t*x)

This can be done by anyone with mathematical knowledge even without knowledge of numpy itself

[Wattsy2020]

Another few ideas test rotation of vectors by the fact that rotating a vector A by 360 degrees shouldn't change it test finding the length of the vector by the fact that the length of a vector A = length of A rotated by any amount

[Wattsy2020]

We also need to test edge cases e.g. Division by 0 is not possible, Matrix inversion should fail for non-square matrices or matrices with a determinant of 0

[Wattsy2020]

I've assigned myself to come up with more linear algebra ideas. Also May to come up with trigonometry ideas e.g. sin(2x) = 2sin(x)cos(x) could be used in a test, as could many other trigonometric identities

[Wattsy2020]

Here's the full list of functionality we need to do to cover most linear algebra strikethrough means completed

Matrices

• Multiplication Rohan
• Addition Rohan
• Trace Anas
• Solving of system of equations Liam
• Transpose Liam
• Pseudo inverse Liam
• Norm of a matrix Liam
• Eigenvector finding Liam
• Eigenvalue finding Liam

Vectors

• Inner product of vectors Parth
• Length of a vector (using np.linalg.norm)