A research poster from the School of Underlying Conditions sets out how many decades of record are needed before a fractional relaxation law can be told from a pair of ordinary exponential decays.
Fractional relaxation, two exponential curves and the record it takes to tell them apart
Some processes forget slowly, and the mathematics of fractional order exists to describe them. Whether a given record has run long enough to show that slow forgetting, rather than something simpler wearing its clothes, is the question a new research poster from the School of Underlying Conditions sets out to answer.
I was taught to let a record run until it had something to say. It is a comfort to have a better idea of when that is, and I hope colleagues find the same comfort in it.
— Dr Vigdis Emenike, Senior Lecturer, School of Underlying Conditions
Working entirely from computed records, the team set a fractional relaxation law against its closest two-term and three-term exponential stand-ins across a spread of orders and noise floors, and located the point at which each stand-in can no longer keep pace. At a noise of one part in a thousand, the poster reports, the simpler account holds for between 1.2 and 2.1 decades of elapsed time, and in some settings no record of any length is enough to unseat it.
The poster joins a body of work the University regards as one of the School’s quieter strengths: patient attention to what a curve can and cannot be asked to carry. Dr Brynjar Obiora, Senior Research Fellow, called it “the kind of result that makes a long record easier to defend and a short one easier to forgive”, while Dr Joost Nwosu, Postdoctoral Fellow, said he had valued “the chance to bring a counting habit to a question about time”.
The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.d5naak.