tvregdiff

Python version of Rick Chartrand's algorithm for numerical differentiation of noisy data. Requires Numpy and Scipy installed. Matplotlib optional for plotting.

Usage:

u = TVRegDiff(data, iter, alph, u0, scale, ep, dx, plotflag, diagflag, precondflag, diffkernel, cgtol, cgmaxit)

Test:

python tvregdiff.py test_data.dat

New parameters

There are a few parameters added with respect to the original script, to allow for greater flexibility. These are:

• precondflag: if set to False, avoid using a preconditioner. Especially for scale='small' problems, sometimes the preconditioner can impede rather than help convergence, and it's useful to turn it off.
• diffkernel: by default is set to 'abs', which means the functional that will be optimised to find the derivative while keeping it smooth depends on the integral of |u'|. However it is also possible to set it to 'sq', which means using instead the integral of (u')^2. In the latter case, the derivative tends to come out smoother, and the need for using more than one iteration is much less. Try which one works best.
• cgtol: tolerance for the conjugate gradient optimisation, previously fixed.
• cgmaxit maximum number of iterations for the conjugate gradient optimisation, previously fixed.

References

Rick Chartrand ([email protected]), Apr. 10, 2011 Please cite Rick Chartrand, "Numerical differentiation of noisy, nonsmooth data," ISRN Applied Mathematics, Vol. 2011, Article ID 164564, 2011.

Algorithm adapted from the Matlab version found here.