Warning Message wrt NMinimize function
in [Mathematica]
|
From: John Adeosun on 21 Sep 2006 08:20 Hello Everyone, I want to use NMinimize to solve an optimization problem. In essence, I'm trying to numerically minimize the "sum of a square of errors" for two lists, aa and bb (earlier generated lists of n elements or points), using the below syntax: NMinimize[{Plus @@ Table[(aa[[i]] - bb[[i]])^2, {i, n}], {y> 0}}, y, Method -> "DifferentialEvolution"] This does not yield the intended result of, say, {minimumvalue, y->10.012}. I have been getting an expression (unevaluated "result" ) with a warning message such as: NMinimize::nnum : The function value {336.381} is not a number at {y}= {0.063703}. More... I then clicked on 'More' for help. I got the below message from the help browser: "General::nnum function::nnum: The function value `3` is not a number at `2` = `1`. Notes for this message have not yet been added." Has anyone experienced this before? Is anything wrong with my syntax above? Thanks in advance for your useful insight. John.
From: Chris Chiasson on 22 Sep 2006 01:26 Maybe you should provide the expression for aa, bb, and n. Also, in regular least squares curve fitting, there is no restriction on the parameter space (no restriction on y). In[1]:= aa=Table[Random[],{20}] Out[1]= {0.303766,0.14977,0.225405,0.0994793,0.358193,0.137617,0.305444,0.0196565,0.\ 418029,0.386865,0.416273,0.492426,0.207767,0.486231,0.426445,0.370501,0.\ 191603,0.921002,0.481863,0.353564} In[2]:= bb=Table[y Random[],{20}] Out[2]= {0.227101 y,0.239582 y,0.0272011 y,0.286646 y,0.923335 y,0.0898122 y, 0.801797 y,0.187167 y,0.565142 y,0.952195 y,0.496352 y,0.16751 y,0.147113 y, 0.56533 y,0.0800794 y,0.675084 y,0.939346 y,0.0790994 y,0.653634 y, 0.304583 y} In[3]:= NMinimize[{Plus@@Table[(aa[[i]]-bb[[i]])^2,{i,Length(a)aa}],{y>0}},y, Method\[Rule]"DifferentialEvolution"] Out[3]= {1.46278,{y\[Rule]0.530985}} On 9/21/06, John Adeosun <jtadeosun(a)yahoo.com> wrote: > Hello Everyone, > > I want to use NMinimize to solve an optimization problem. In essence, I'm > trying to numerically minimize the "sum of a square of errors" for two > lists, aa and bb (earlier generated lists of n elements or points), using > the below syntax: > > NMinimize[{Plus @@ Table[(aa[[i]] - bb[[i]])^2, {i, n}], {y> 0}}, y, > Method -> "DifferentialEvolution"] > This does not yield the intended result of, say, {minimumvalue, y->10.012}. > I have been getting an expression (unevaluated "result" ) with a warning > message such as: > > NMinimize::nnum : The function value {336.381} is not a number at {y}= > {0.063703}. More... > > I then clicked on 'More' for help. I got the below message from the help > browser: > > "General::nnum > > function::nnum: The function value `3` is not a number at `2` = `1`. > > Notes for this message have not yet been added." > > Has anyone experienced this before? Is anything wrong with my syntax above? > > Thanks in advance for your useful insight. > > John. > > > > > > > -- http://chris.chiasson.name/
From: Jens-Peer Kuska on 22 Sep 2006 01:38 Hi, regardless of the fact that your input is nonsense, because your objective function does not depend on y at all With[{n = 4}, aa = Table[Random[], {n}]; bb = Table[Random[], {n}]; NMinimize[{Plus @@ Table[(aa[[i]] - bb[[i]])^2, {i, n}], {y > 0}}, y, Method -> "DifferentialEvolution"] ] work fine for me and I assume, that you hide some facts to prevent us from answer your question. Regards Jens "John Adeosun" <jtadeosun(a)yahoo.com> schrieb im Newsbeitrag news:eeu01h$b83$1(a)smc.vnet.net... | Hello Everyone, | | I want to use NMinimize to solve an optimization problem. In essence, I'm | trying to numerically minimize the "sum of a square of errors" for two | lists, aa and bb (earlier generated lists of n elements or points), using | the below syntax: | | NMinimize[{Plus @@ Table[(aa[[i]] - bb[[i]])^2, {i, n}], {y> 0}}, y, | Method -> "DifferentialEvolution"] | This does not yield the intended result of, say, {minimumvalue, y->10.012}. | I have been getting an expression (unevaluated "result" ) with a warning | message such as: | | NMinimize::nnum : The function value {336.381} is not a number at {y}= | {0.063703}. More... | | I then clicked on 'More' for help. I got the below message from the help | browser: | | "General::nnum | | function::nnum: The function value `3` is not a number at `2` = `1`. | | Notes for this message have not yet been added." | | Has anyone experienced this before? Is anything wrong with my syntax above? | | Thanks in advance for your useful insight. | | John. | | | | | |
From: Jean-Marc Gulliet on 22 Sep 2006 01:40 John Adeosun wrote: > Hello Everyone, > > I want to use NMinimize to solve an optimization problem. In essence, I'm > trying to numerically minimize the "sum of a square of errors" for two > lists, aa and bb (earlier generated lists of n elements or points), using > the below syntax: > > NMinimize[{Plus @@ Table[(aa[[i]] - bb[[i]])^2, {i, n}], {y> 0}}, y, > Method -> "DifferentialEvolution"] > This does not yield the intended result of, say, {minimumvalue, y->10.012}. > I have been getting an expression (unevaluated "result" ) with a warning > message such as: > > NMinimize::nnum : The function value {336.381} is not a number at {y}= > {0.063703}. More... > > I then clicked on 'More' for help. I got the below message from the help > browser: > > "General::nnum > > function::nnum: The function value `3` is not a number at `2` = `1`. > > Notes for this message have not yet been added." > > Has anyone experienced this before? Is anything wrong with my syntax above? John, The syntax above seems correct (see below). I suspect you have some problem with your lists. Note that the second version of the code should be more efficient (you can do vector operations). In[1]:= aa = Table[Random[], {10}]; bb = Table[Random[], {10}]; In[3]:= NMinimize[{Plus @@ Table[(aa[[i]] - bb[[i]])^2, {i, 10}], {y > 0}}, y, Method -> "DifferentialEvolution"] Out[3]= {1.71356,{y\[Rule]13.0914}} In[4]:= NMinimize[{Plus @@ ((aa - bb)^2), {y > 0}}, y, Method -> "DifferentialEvolution"] Out[4]= {1.71356,{y\[Rule]13.0914}} In[5]:= $Version Out[5]= 5.2 for Microsoft Windows (June 20, 2005) HTH, Jean-Marc
From: p-valko on 22 Sep 2006 01:42 As it is written right now, you are minimizing a list that consists of two elements, each of them being a list itself. NMinimize really does not know what to do with such an object. (Very often in Mathematica - and in any software - the error messages are difficult to understand because they are logical only within a given context, but the user misses the context itself, not the particulars within the context. I think the next generation of softwares will have some AI built-in to recognize the real root of the misunderstanding and will offer more help. Today softwares, like Mathematica, still look like a bit autistic persons, but soon they will have more empathy for the user. ) John Adeosun wrote: > Hello Everyone, > > I want to use NMinimize to solve an optimization problem. In essence, I'm > trying to numerically minimize the "sum of a square of errors" for two > lists, aa and bb (earlier generated lists of n elements or points), using > the below syntax: > > NMinimize[{Plus @@ Table[(aa[[i]] - bb[[i]])^2, {i, n}], {y> 0}}, y, > Method -> "DifferentialEvolution"] > This does not yield the intended result of, say, {minimumvalue, y->10.012}. > I have been getting an expression (unevaluated "result" ) with a warning > message such as: > > NMinimize::nnum : The function value {336.381} is not a number at {y}= > {0.063703}. More... > > I then clicked on 'More' for help. I got the below message from the help > browser: > > "General::nnum > > function::nnum: The function value `3` is not a number at `2` = `1`. > > Notes for this message have not yet been added." > > Has anyone experienced this before? Is anything wrong with my syntax above? > > Thanks in advance for your useful insight. > > John.
|
Pages: 1 Prev: (again) common problem with FindMinimum and NIntegrate Next: Possible simple bug in NMaximize |