This is an archival dump of old wiki content --- see for current material

Development of a new Polynomial class for Scipy


Currently numpy provides a class, poly1d, for representing polynomials in one variable. This class always represents the polynomials as arrays of coefficients in the basis 1, x, x**2, ..., x**n. Unfortunately this basis yields very poor numerical representations of many useful polynomials. As a result, if users want to work with polynomial objects, they are often stuck using low orders. For example, scipy.special includes several families of orthogonal polynomials described in terms of recurrence relations. The objects returned when one is created are poly1ds, and as a result the Chebyshev polynomials become useless by degree 35, while the underlying recurrence relations, with an appropriate evaluation scheme, would allow much higher degrees.


Use cases



A prototype is available on github as scikits.polynomial.


(Mostly behind paywalls I'm afraid.)

SciPy: NewPolynomials (last edited 2015-10-24 17:48:24 by anonymous)