Hi, First of all, this mailing list is shared among all commons projects. Please add a marker like [math] in the subject line when posting to it to help subscribers filter messages. I have added it for the reply.
----- "hofi85" <[email protected]> a écrit : > hi there, > > i was looking for numerical differentiation routines in the api > (forward, > backward central difference rules) but i can't find them. am i just > looking > in the wrong packages (analysis)? can you point me to right class/ > method? There is no differentiation feature yet in commons-math. A separate project has been started on the subject : commons-nabla. This project is still in development. I have put it on hold temporarily since I ma busy working on version 2.0 of commons-math. Work will be restarted on commons-nabla once 2.0 is out. Commons-nabla is a sandbox project, so there is no realease yet. You can browse the site here: http://commons.apache.org/sandbox/nabla/ and you can checkout the project source from subversion here: http://svn.apache.org/repos/asf/commons/sandbox/nabla/trunk. The goal of commons-nabla is to provide automatic differentiation directly from Java classes. It defines an UnivariateDifferentiable interface which must be implemented by your function and an UnivariateDifferentiator interface to transform your function into another object computing the derivative. The component provides two implementations of the UnivariateDifferentiator interface. One is already usable but is based on finite differences and hence has some drawbacks (approximate computation, needs a step size, computation cost ...). Another one is not finished yet (it can handle only very simple functions yet) but is based on exact differentiation by bytecode analysis, differentiation and new bytecode generation on the fly. Of course, if you want to use commons-nabla but encounters an issue, you can register a ticket in Jira (see the link "Issue Tracking" in the menu on nabla site). There are currently no tickets at all for this component. regards, Luc > > thx > > u --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
