My Math Forum Diagnostics to test divergence of an unknown function

 Computer Science Computer Science Forum

 October 8th, 2015, 07:45 AM #1 Senior Member   Joined: Apr 2014 From: Glasgow Posts: 2,157 Thanks: 732 Math Focus: Physics, mathematical modelling, numerical and computational solutions Diagnostics to test divergence of an unknown function I have a problem associated with convergence in a computer program. Consider a numerical solution based on iteration with the form $\displaystyle x_{n+1} = w f(x_n) + (1-w) x_n$ where $\displaystyle w$ is known as the relaxation factor. Basically, you mix a portion of the original answer together with the new result using a weighting factor. By setting $\displaystyle 0 \lt w \le 1$ (called under-relaxation) you can improve the stability of the solution at the cost of convergence rate. The following properties are true: 1. $\displaystyle f(x_n)$ is unknown, as are its derivatives. However, we can be sure that $\displaystyle f(x_n)$ does not depend on $\displaystyle n$ and is "sensible" (no discontinuities, single-valued over the reals, etc.) 2. A solution indeed exists such that after some threshold number of iterations, $\displaystyle n > N$, and convergent iteration,$\displaystyle \frac{x_{N+1} - x_N}{x_N} < \delta$ where $\displaystyle \delta$ is some sufficiently small number (such as 0.01) is true. Does anyone know of a decent algorithm for detecting divergence and, if so, how to set the under-relaxation factor to compensate?

 Tags diagnostics, divergence, function, test, unknown

 Thread Tools Display Modes Linear Mode

 Similar Threads Thread Thread Starter Forum Replies Last Post king.oslo Complex Analysis 6 May 22nd, 2015 02:00 PM mathdisciple Calculus 6 March 23rd, 2014 12:24 AM hatsjoe Linear Algebra 0 March 17th, 2014 06:28 AM shiseonji Calculus 1 February 21st, 2014 04:53 AM aaron-math Calculus 8 October 12th, 2011 10:03 PM

 Contact - Home - Forums - Cryptocurrency Forum - Top