login  home  contents  what's new  discussion  bug reports     help  links  subscribe  changes  refresh  edit

Edit detail for SandBoxNewtonsMethod revision 2 of 2

1 2
Editor: Mark Clements
Time: 2009/04/26 23:45:04 GMT-7
Note: Included Waldek Hebisch's suggestion for mkLispFunction1

changed:
-)lisp (defun newton (f dfdx x0 &optional (tol 1.0d-10)) (let ((xt x0) (fxt (funcall f x0)) (maxit 100) (iternum 0)) (loop (setf xt (- xt (/ fxt (funcall dfdx xt)))) (setf fxt (funcall f xt)) (if (< (abs fxt) tol) (return xt)) (incf iternum) (if (>= iternum maxit) (error "Maximum iterations exceeded.")))))
)lisp (defun newton (f dfdx x0 &optional (tol 1.0d-10) (maxit 100)) (let ((xt x0) (fxt (funcall f x0)) (iternum 0)) (loop (setf xt (- xt (/ fxt (funcall dfdx xt)))) (setf fxt (funcall f xt)) (if (< (abs fxt) tol) (return xt)) (incf iternum) (if (>= iternum maxit) (error "Maximum iterations exceeded.")))))

added:

Finally, within a Spad package, it is even easier to call a function through Lisp:

\begin{spad}
-- the following approach is due to Waldek Hebisch
)lisp (defun |mkLispFunction1| (f) (lambda (x) (SPADCALL x f)))
)abbrev package TESTP TestPackage
R ==> DoubleFloat
I ==> Integer
TestPackage: with
    newton:(R->R,R->R,R,R,I)->R
    newton:(R->R,R->R,R)->R
  == add
    newton(f,dfdx,x0,tol,maxit) ==
         NEWTON(mkLispFunction1(f@(R -> R))$Lisp,
                 mkLispFunction1(dfdx@(R -> R))$Lisp,
                 x0,tol,maxit)$Lisp
    newton(f,dfdx,x0) == newton(f,dfdx,x0,1.0e-10::R,100::I)
\end{spad}

then

\begin{axiom}
R ==> DoubleFloat
newton((x:R):R +-> x**2-2.0::R,(x:R):R +-> 2.0::R*x,3.0::SF,1.0e-15,100)$TestPackage-sqrt(2.0)
newtonExpression(f,x,x0) ==
	dfdx := D(f,x)
	newton((xt:R):R +-> eval(f,x,xt), (xt:R):R +-> eval(dfdx,x,xt), x0)$TestPackage
newtonExpression(x**2-2.0,x,2.0)-sqrt(2.0)
\end{axiom}

The following shows Newton's method for numerically solving f(x)=0. It is also shows examples of calling Axiom expressions and Spad functions from Lisp.

First, define Newton's method using the Axiom interpreter:

fricas
-- Axiom interpreter function for Newton's algorithm
R ==> Float
Type: Void
fricas
I ==> Integer
Type: Void
fricas
newton(f:Expression R,x:Symbol,x0:R):R ==
        dfdx:Expression R := D(f,x)
        xt:R := x0
        fxt:R := subst(f,x=xt)
        iterNum:I := 0
        maxIt:I := 100
        repeat
                xt := xt - fxt/subst(dfdx,x=xt)
                fxt := subst(f,x=xt)
                if abs(fxt)<1.0e-10 then return xt
                iterNum:=iterNum+1::I
                if iterNum >= maxIt then 
                        error "Maximum iterations exceeded."
Function declaration newton : (Expression(Float), Symbol, Float) -> Float has been added to workspace.
Type: Void
fricas
newton(x**2-2.0,x,2.0)
There are no library operations named ** Use HyperDoc Browse or issue )what op ** to learn if there is any operation containing " ** " in its name.
Cannot find a definition or applicable library operation named ** with argument type(s) Expression(Float) PositiveInteger
Perhaps you should use "@" to indicate the required return type, or "$" to specify which version of the function you need.

Second, we can also do this by calling the Newton method implemented in Lisp. We initially define/hack a Spad package which creates a Lisp function from an interpreter expression.

spad
)lisp (defun |lambdaFuncallSpad| (f) (lambda (x) (funcall f x nil)))
)abbrev package MKULF MakeUnaryLispFunction
++ Tools for making compiled lisp functions from top-level expressions
++ Author: Mark Clements 
++ (based on the MakeUnaryCompiledFunction package by Manuel Bronstein)
++ Date Created: 20 April 2008
++ Date Last Updated: 20 April 2008
++ Description: transforms top-level objects into lisp functions.
MakeUnaryLispFunction(S, D, I): Exports == Implementation where
  S: ConvertibleTo InputForm
  D, I: Type
SY ==> Symbol DI ==> devaluate(D -> I)$Lisp
Exports ==> with compiledFunction: (S, SY) -> SY ++ compiledFunction(expr, x) returns a function lisp{f: D -> I} ++ defined by lisp{(defun f (x) expr)}. ++ Function f is compiled and directly ++ applicable to objects of type D (in lisp).
Implementation ==> add import MakeFunction(S)
compiledFunction(e:S, x:SY) == t := [convert([devaluate(D)$Lisp]$List(InputForm)) ]$List(InputForm) lambdaFuncallSpad(compile(function(e, declare DI, x), t))$Lisp
spad
   Compiling FriCAS source code from file 
      /var/lib/zope2.10/instance/axiom-wiki/var/LatexWiki/1657093691659194492-25px002.spad
      using old system compiler.
Value = |lambdaFuncallSpad|
   MKULF abbreviates package MakeUnaryLispFunction 
------------------------------------------------------------------------
   initializing NRLIB MKULF for MakeUnaryLispFunction 
   compiling into NRLIB MKULF 
   importing MakeFunction S
   compiling exported compiledFunction : (S,Symbol) -> Symbol
Time: 0 SEC.
(time taken in buildFunctor: 0)
;;; *** |MakeUnaryLispFunction| REDEFINED
;;; *** |MakeUnaryLispFunction| REDEFINED Time: 0 SEC.
Cumulative Statistics for Constructor MakeUnaryLispFunction Time: 0 seconds
finalizing NRLIB MKULF Processing MakeUnaryLispFunction for Browser database: --------constructor--------- --------(compiledFunction ((Symbol) S (Symbol)))--------- ; compiling file "/var/aw/var/LatexWiki/MKULF.NRLIB/MKULF.lsp" (written 04 APR 2022 07:16:37 PM):
; /var/aw/var/LatexWiki/MKULF.NRLIB/MKULF.fasl written ; compilation finished in 0:00:00.014 ------------------------------------------------------------------------ MakeUnaryLispFunction is now explicitly exposed in frame initial MakeUnaryLispFunction will be automatically loaded when needed from /var/aw/var/LatexWiki/MKULF.NRLIB/MKULF

Then we can use this in Axiom...

fricas
-- define Newton in Lisp
fricas
)lisp (defun newton (f dfdx x0 &optional (tol 1.0d-10) (maxit 100)) (let ((xt x0) (fxt (funcall f x0)) (iternum 0)) (loop (setf xt (- xt (/ fxt (funcall dfdx xt)))) (setf fxt (funcall f xt)) (if (< (abs fxt) tol) (return xt)) (incf iternum) (if (>= iternum maxit) (error "Maximum iterations exceeded.")))))
Value = NEWTON -- Spad->Lisp function translation compiledDF(expr: EXPR FLOAT, x: Symbol):Symbol == compiledFunction(expr,x)$MakeUnaryLispFunction(EXPR FLOAT,DFLOAT,DFLOAT)
Function declaration compiledDF : (Expression(Float), Symbol) -> Symbol has been added to workspace.
Type: Void
fricas
-- a short function to call the Lisp code
newtonUsingLisp(f:Expression Float,x:Symbol,x0:DFLOAT):DFLOAT ==
        float(NEWTON(compiledDF(f,x),compiledDF(D(f,x),x),x0)$Lisp)
Function declaration newtonUsingLisp : (Expression(Float), Symbol, DoubleFloat) -> DoubleFloat has been added to workspace.
Type: Void
fricas
newtonUsingLisp(x**2-2.0,x,2.0::SF)-sqrt(2.0::SF)
There are no library operations named ** Use HyperDoc Browse or issue )what op ** to learn if there is any operation containing " ** " in its name.
Cannot find a definition or applicable library operation named ** with argument type(s) Expression(Float) PositiveInteger
Perhaps you should use "@" to indicate the required return type, or "$" to specify which version of the function you need.

Third, we can call a compiled Spad function in Lisp. Defining some example functions in Spad:

spad
)abbrev package TESTP TestPackage
R ==> DoubleFloat
TestPackage: with
        f:R -> R
        dfdx:R -> R
  == add
        f(x) == x*x - 2.0::R
        dfdx(x) == 2*x
spad
   Compiling FriCAS source code from file 
      /var/lib/zope2.10/instance/axiom-wiki/var/LatexWiki/2983317606354902709-25px004.spad
      using old system compiler.
   TESTP abbreviates package TestPackage 
------------------------------------------------------------------------
   initializing NRLIB TESTP for TestPackage 
   compiling into NRLIB TESTP 
   compiling exported f : DoubleFloat -> DoubleFloat
Time: 0.01 SEC.
compiling exported dfdx : DoubleFloat -> DoubleFloat Time: 0.01 SEC.
(time taken in buildFunctor: 0)
;;; *** |TestPackage| REDEFINED
;;; *** |TestPackage| REDEFINED Time: 0 SEC.
Cumulative Statistics for Constructor TestPackage Time: 0.02 seconds
finalizing NRLIB TESTP Processing TestPackage for Browser database: --->-->TestPackage(constructor): Not documented!!!! --->-->TestPackage((f ((DoubleFloat) (DoubleFloat)))): Not documented!!!! --->-->TestPackage((dfdx ((DoubleFloat) (DoubleFloat)))): Not documented!!!! --->-->TestPackage(): Missing Description ; compiling file "/var/aw/var/LatexWiki/TESTP.NRLIB/TESTP.lsp" (written 04 APR 2022 07:16:37 PM):
; /var/aw/var/LatexWiki/TESTP.NRLIB/TESTP.fasl written ; compilation finished in 0:00:00.009 ------------------------------------------------------------------------ TestPackage is now explicitly exposed in frame initial TestPackage will be automatically loaded when needed from /var/aw/var/LatexWiki/TESTP.NRLIB/TESTP

and then calling those functions in Lisp from Axiom:

fricas
)lisp (defun |lispFunctionFromSpad| (f dom args) (let ((spadf (|getFunctionFromDomain| f (list dom) args))) (lambda (x) (spadcall x spadf))))
Value = |lispFunctionFromSpad| float(NEWTON(lispFunctionFromSpad(f,'TestPackage,['DoubleFloat])$Lisp, lispFunctionFromSpad(dfdx,'TestPackage,['DoubleFloat])$Lisp,2::SF)$Lisp)-sqrt(2.0::SF)

\label{eq1}1.5947243525715749 e - 12(1)
Type: DoubleFloat?

Finally, within a Spad package, it is even easier to call a function through Lisp:

spad
-- the following approach is due to Waldek Hebisch
)lisp (defun |mkLispFunction1| (f) (lambda (x) (SPADCALL x f)))
)abbrev package TESTP TestPackage
R ==> DoubleFloat
I ==> Integer
TestPackage: with
    newton:(R->R,R->R,R,R,I)->R
    newton:(R->R,R->R,R)->R
  == add
    newton(f,dfdx,x0,tol,maxit) ==
         NEWTON(mkLispFunction1(f@(R -> R))$Lisp,
                 mkLispFunction1(dfdx@(R -> R))$Lisp,
                 x0,tol,maxit)$Lisp
    newton(f,dfdx,x0) == newton(f,dfdx,x0,1.0e-10::R,100::I)
spad
   Compiling FriCAS source code from file 
      /var/lib/zope2.10/instance/axiom-wiki/var/LatexWiki/5566883931090623725-25px006.spad
      using old system compiler.
Value = |mkLispFunction1|
   TESTP abbreviates package TestPackage 
------------------------------------------------------------------------
   initializing NRLIB TESTP for TestPackage 
   compiling into NRLIB TESTP 
   compiling exported newton : (DoubleFloat -> DoubleFloat,DoubleFloat -> DoubleFloat,DoubleFloat,DoubleFloat,Integer) -> DoubleFloat
Time: 0 SEC.
compiling exported newton : (DoubleFloat -> DoubleFloat,DoubleFloat -> DoubleFloat,DoubleFloat) -> DoubleFloat Time: 0.01 SEC.
(time taken in buildFunctor: 0)
;;; *** |TestPackage| REDEFINED
;;; *** |TestPackage| REDEFINED Time: 0 SEC.
Cumulative Statistics for Constructor TestPackage Time: 0.01 seconds
finalizing NRLIB TESTP Processing TestPackage for Browser database: --->/var/lib/zope2.10/instance/axiom-wiki/var/LatexWiki/2983317606354902709-25px004.spad-->TestPackage(constructor): Not documented!!!! --->/var/lib/zope2.10/instance/axiom-wiki/var/LatexWiki/2983317606354902709-25px004.spad-->TestPackage((newton ((DoubleFloat) (Mapping (DoubleFloat) (DoubleFloat)) (Mapping (DoubleFloat) (DoubleFloat)) (DoubleFloat) (DoubleFloat) (Integer)))): Not documented!!!! --->/var/lib/zope2.10/instance/axiom-wiki/var/LatexWiki/2983317606354902709-25px004.spad-->TestPackage((newton ((DoubleFloat) (Mapping (DoubleFloat) (DoubleFloat)) (Mapping (DoubleFloat) (DoubleFloat)) (DoubleFloat)))): Not documented!!!! --->/var/lib/zope2.10/instance/axiom-wiki/var/LatexWiki/2983317606354902709-25px004.spad-->TestPackage(): Missing Description ; compiling file "/var/aw/var/LatexWiki/TESTP.NRLIB/TESTP.lsp" (written 04 APR 2022 07:16:37 PM):
; /var/aw/var/LatexWiki/TESTP.NRLIB/TESTP.fasl written ; compilation finished in 0:00:00.009 ------------------------------------------------------------------------ TestPackage is already explicitly exposed in frame initial TestPackage will be automatically loaded when needed from /var/aw/var/LatexWiki/TESTP.NRLIB/TESTP

then

fricas
R ==> DoubleFloat
Type: Void
fricas
newton((x:R):R +-> x**2-2.0::R,(x:R):R +-> 2.0::R*x,3.0::SF,1.0e-15,100)$TestPackage-sqrt(2.0)
There are no library operations named ** Use HyperDoc Browse or issue )what op ** to learn if there is any operation containing " ** " in its name. Cannot find a definition or applicable library operation named ** with argument type(s) DoubleFloat PositiveInteger
Perhaps you should use "@" to indicate the required return type, or "$" to specify which version of the function you need. FriCAS will attempt to step through and interpret the code. There are no library operations named ** Use HyperDoc Browse or issue )what op ** to learn if there is any operation containing " ** " in its name.
Cannot find a definition or applicable library operation named ** with argument type(s) Symbol PositiveInteger
Perhaps you should use "@" to indicate the required return type, or "$" to specify which version of the function you need.