Advanced Search
All Posts : Product Tips & Techniques
As currently programmed, fsolve() does not do numerical derivatives for systems of equations. The reason for this is that subs() is used instead of eval() when evaluating derivatives. [Note: jacob is the symbolically defined jacobian of the system of equations and lsub is the sequence of appropriate numerical substitutions (not the "list" of substitutions the mnemonic might suggest).] The original statement (in `fsolve/sysnewton`) is: A:=traperror(evalf(subs(lsub,jacob),Digits-5+n)); The functions evalf() and subs() apparently do no know how to work together to produce numerical derivatives. However, the following statement does work. A:=traperror(evalf(eval(jacob,[lsub]),Digits-5+n)); This combination of evalf() and eval() recognizes the need for a numerical derivative and sets up a call to fdiff(). See the help page for fdiff().

Please Wait...