Abstract
When representing the coefficients and knots of a spline using only small integers, independently rounding each infinite-precision value is not the best strategy. We show how to build an affine model for the error expanded about the optimal full-precision free-knot or parameterized spline, then use the Lovasz basis reduction algorithm to select a better rounding. The technique could be used for other situations in which a quadratic error model can be computed.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have