2 * Introduction to hompack::
3 * Functions and Variables for hompack::
6 @node Introduction to hompack, Functions and Variables for hompack
7 @section Introduction to hompack
9 @code{Hompack} is a Common Lisp translation (via @code{f2cl}) of the
10 Fortran library HOMPACK, as obtained from Netlib.
12 @opencatbox{Categories:}
13 @category{Numerical methods}
14 @category{Share packages}
15 @category{Package hompack}
18 @node Functions and Variables for hompack, , Introduction to hompack
19 @section Functions and Variables for hompack
21 @anchor{hompack_polsys}
22 @deffn {Function} hompack_polsys (@var{eqnlist}, @var{varlist} [, @var{iflg1}, @var{epsbig}, @var{epssml}, @var{numrr}])
23 Finds the roots of the system of polynomials in the variables
24 @var{varlist} in the system of equations in @var{eqnlist}. The number
25 of equations must match number of variables. Each equation must be a
26 polynomial with variables in @var{varlist}. The coefficients must be
29 The optional keyword arguments provide some control over the
30 algorithm. @var{epsbig} is the local error tolerance allowed by the
31 path tracker, defaulting to 1e-4. @var{epssml} is the accuracy
32 desired for the final solution, defaulting to 1d-14.. @var{iflg1},
33 defaulting to 0, controls the algorithm as follows:
36 If the problem is to be solved without calling @code{polsys}' scaling
37 routine, @code{sclgnp}, and without using the projective
40 If scaling but no projective transformation is to be used.
42 If no scaling but projective transformation is to be used.
44 If both scaling and projective transformation are to be used.
47 @code{hompack_polsys} returns a list. The elements of the list are:
50 Indicates whether the solution is valid or not.
53 Solution found without problems
55 Solution succeeded but @code{iflg2} indicates some issues with a
56 root. (That is, @code{iflg2} is not all ones.)
58 @code{NN}, the declared dimension of the number of terms in the
59 polynomials, is too small. (This should not happen.)
61 @code{MMAXT}, the declared dimension for the internal coefficient and
62 degree arrays, is too small. (This should not happen.)
64 @code{TTOTDG}, the total degree of the equations, is too small.
65 (This should not happen.)
67 @code{LENWK}, the length of the internal real work array, is too
68 small. (This should not happen.)
70 @code{LENIWK}, the length of the internal integer work array, is too
71 small. (This should not happen.)
73 @var{iflg1} is not 0 or 1, or 10 or 11. (This should not happen; an
74 error should be thrown before @code{polsys} is called.)
77 The roots of the system of equations. This is in the same format as
78 @code{solve} would return.
80 Information on how the path for the m'th root terminated:
85 Specified error tolerance cannot be met. Increase @var{epsbig} and
86 @var{epssml} and rerun.
88 Maximum number of steps exceeded. To track the path further, increase
89 @var{numrr} and rerun the path. However, the path may be diverging, if the
90 lambda value is near 1 and the roots values are large.
92 Jacobian matrix does not have full rank. The algorithm has failed
93 (the zero curve of the homotopy map cannot be followed any further).
95 The tracking algorithm has lost the zero curve of the homotopy map and
96 is not making progress. The error tolerances @var{epsbig} and
97 @var{epssml} were too lenient. The problem should be restarted with
98 smaller error tolerances.
100 The normal flow newton iteration in @code{stepnf} or @code{rootnf}
101 failed to converge. The error tolerance @var{epsbig} may be too
104 Illegal input parameters, a fatal error.
107 The final lambda value for the m-th root, where lambda is the
108 continuation parameter.
110 The arc length of the m-th root.
112 The number of jacobian matrix evaluations required to track the m-th
117 (%i1) hompack_polsys([x1^2-1, x2^2-2],[x1,x2]);
119 [[x1 = (-1.354505666901954e-16*%i)-0.9999999999999999,
120 x2 = 3.52147935979316e-16*%i-1.414213562373095],
121 [x1 = 1.0-5.536432658639868e-18*%i,
122 x2 = (-4.213674137126362e-17*%i)-1.414213562373095],
123 [x1 = (-9.475939894034927e-17*%i)-1.0,
124 x2 = 2.669654624736742e-16*%i+1.414213562373095],
125 [x1 = 9.921253413273088e-18*%i+1.0,
126 x2 = 1.414213562373095-5.305667769855424e-17*%i]],[1,1,1,1],
127 [1.0,1.0,0.9999999999999996,1.0],
128 [4.612623769341193,4.612623010859902,4.612623872939383,
129 4.612623114484402],[40,40,40,40]]
132 The analytical solution can be obtained with solve:
134 (%i2) solve([x1^2-1, x2^2-2],[x1,x2]);
135 (%o2) [[x1 = 1,x2 = -sqrt(2)],[x1 = 1,x2 = sqrt(2)],[x1 = -1,x2 = -sqrt(2)],
136 [x1 = -1,x2 = sqrt(2)]]
138 We see that @code{hompack_polsys} returned the correct answer except
139 that the roots are in a different order and there is a small imaginary
142 Another example, with corresponding solution from solve:
144 (%i1) hompack_polsys([x1^2 + 2*x2^2 + x1*x2 - 5, 2*x1^2 + x2^2 + x2-4],[x1,x2]);
146 [[x1 = 1.201557301700783-1.004786320788336e-15*%i,
147 x2 = (-4.376615092392437e-16*%i)-1.667270363480143],
148 [x1 = 1.871959754090949e-16*%i-1.428529189565313,
149 x2 = (-6.301586314393093e-17*%i)-0.9106199083334113],
150 [x1 = 0.5920619420732697-1.942890293094024e-16*%i,
151 x2 = 6.938893903907228e-17*%i+1.383859154368197],
152 [x1 = 7.363503717463654e-17*%i+0.08945540033671608,
153 x2 = 1.557667481081721-4.109128293931921e-17*%i]],[1,1,1,1],
154 [1.000000000000001,1.0,1.0,1.0],
155 [6.205795654034752,7.722213259390295,7.228287079174351,
156 5.611474283583368],[35,41,48,40]]
157 (%i2) solve([x1^2+2*x2^2+x1*x2 - 5, 2*x1^2+x2^2+x2-4],[x1,x2]);
158 (%o2) [[x1 = 0.08945540336850383,x2 = 1.557667386609071],
159 [x1 = 0.5920619554695062,x2 = 1.383859286083807],
160 [x1 = 1.201557352500749,x2 = -1.66727025803531],
161 [x1 = -1.428529150636283,x2 = -0.9106198942815954]]
164 Note that @code{hompack_polsys} can sometimes be very slow. Perhaps
165 @code{solve} can be used. Or perhaps @code{eliminate} can be used to
166 convert the system of polynomials into one polynomial for which
167 @code{allroots} can find all the roots.
172 @c TeX-master: "include-maxima"