Eigenvectors and eigenvalues
Not met yet
Multiply a direction by a matrix and it usually turns. An eigenvector is a direction that doesn't: matrix @ v comes out as value * v, the same arrow stretched by its eigenvalue. For a covariance matrix, the eigenvectors are the directions the readings spread along, and each eigenvalue is the variance along its direction. values, vectors = np.linalg.eigh(matrix), for symmetric matrices like covariances, gives the eigenvalues smallest first, and each eigenvector as a column: vectors[:, j] goes with values[j]. An eigenvector and its negative are the same line, so eigh may hand back either.
Try it
main.pyOutput
Press Run and what your program prints shows up here.
Easy to get wrong
- Thinks np.linalg.eigh hands back the eigenvalues biggest first, rather than smallest first.
- Thinks the eigenvalues of a matrix are the numbers on its diagonal.
- Thinks np.linalg.eigh gives each eigenvector as a row, rather than as a column.
- Thinks an eigenvector and its negative are different directions, so the sign eigh gives means something.
- Thinks multiplying by a matrix turns every direction, so no arrow keeps pointing the same way.