On Mon, Jan 13, 2020 at 03:04:55PM +1100, Valeri Aronov wrote: > Dear all, > > I trust that I have an excellent proposition for the theme in the subject > of this post. Using my algorithm for the evaluation of all the first and > second derivatives (by circuit parameters) of the symbolic transfer > function asymptotically approaches the time of the evaluation of transfer > function itself. > > Using these derivatives allows substantially reduce the time of circuit > parameter optimization because of much better precision of the derivatives Sounds interesting for SPICE as "Simulation Program with Integrated Circuit Emphasis" [0] while this is a different kind of SPICE list [1] [0] https://en.wikipedia.org/wiki/SPICE [1] https://en.wikipedia.org/wiki/Simple_Protocol_for_Independent_Computing_Environments Cheers, Victor > evaluation (in comparison with non-symbolic derivatives evaluation) and an > opportunity to apply optimization methods using the second derivatives. > > I am looking for discussing with or presenting this to ppl who implemented > symbolic transfer function generation or interested in using the symbolic > functions for parameter optimization or sensitivity evaluation.
Attachment:
signature.asc
Description: PGP signature
_______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/spice-devel