Re: [ojAlgo-user] QuadraticSolver issue

[Anders P. <an...@op...>]

Here's what I get:

[AE] = class org.ojalgo.matrix.PrimitiveMatrix
{{1.0,	1.0,	1.0}}

[BE] = class org.ojalgo.matrix.PrimitiveMatrix
{{1.0}}

[Q] = class org.ojalgo.matrix.PrimitiveMatrix
{{4.0,	1.0,	2.0},
{1.0,	9.0,	1.0},
{2.0,	1.0,	16.0}}

[C] = class org.ojalgo.matrix.PrimitiveMatrix
{{10.0},
{15.0},
{18.0}}

[AI] = class org.ojalgo.matrix.PrimitiveMatrix
{{1.0,	0.0,	0.0},
{0.0,	1.0,	0.0},
{0.0,	0.0,	1.0}}

[BI] = class org.ojalgo.matrix.PrimitiveMatrix
{{1.0},
{1.0},
{1.0}}

[X] = class org.ojalgo.matrix.PrimitiveMatrix
{{-0.17441860465116266},
{0.6279069767441859},
{0.5465116279069767}}

[LE] = class org.ojalgo.matrix.PrimitiveMatrix
{{8.976744186046512}}

[LI] = class org.ojalgo.matrix.PrimitiveMatrix
{{0.0},
{0.0},
{0.0}}

[SE] = class org.ojalgo.matrix.PrimitiveMatrix
{{0.0}}

[SI] = class org.ojalgo.matrix.PrimitiveMatrix
{{1.1744186046511627},
{0.37209302325581406},
{0.4534883720930233}}

I do not get any weights larger than 1.0, but I do get one weight smaller than 0.0. There's no constraint to restrict that! Add constraints for that and you should get the solutions you expect.

Here's how I would find the efficient frontier:

        final BasicMatrix tmpCovars = ...;
        final BasicMatrix tmpReturs = ...;

        final MarketEquilibrium tmpME = new MarketEquilibrium(tmpCovars);

        final MarkowitzModel tmpMarkowitz = new MarkowitzModel(tmpME, tmpReturs);

        for (int i = 1; i < 10; i++) { // Some different risk aversion factors

            tmpMarkowitz.setRiskAversion(new BigDecimal(i));

            final List<BigDecimal> tmpWeights = tmpMarkowitz.getWeights();
        }





On 14 jun 2011, at 17.59, Simon Greaves wrote:

> Well I made this test case, but it requires a utilities library to load the files that contain the matrices.
> You can see what I'm doing but it can't run.
> I'm trying to to Efficient Frontier analysis. I tried using the ojAlgo.finance.porfolio.MarkowitzModel class, but I had a lot of trouble understanding the doc file. So I decided to to it myself using the QuadraticSolver, which I had less trouble with...
> 
> I'm solving (as in the doc file):
> min 1/2 [X]T[Q][X] - [C]T[X]
> when [AE][X] == [BE]
> and [AI][X] <= [BI]
> 
> where (in the test case)
> Q (Covariance Matrix):
> 0.04	0.01	0.02
> 0.01	0.09	0.01
> 0.02	0.01	0.16
> 
> C (Expected Return Vector):
> 0.10
> 0.15
> 0.18
> 
> The sum of X should be 1 since it's portfolio analysis, and I should use 100% of the investment on all activities.
> And every Xi should be between 0 and 1.
> So I made AE a vector of 1's and BE=1, so that takes care of sum(X)=1. And
> AI is an identity matrix, and BI a 1 vector. So all Xi's should be <=1.
> 
> But as a result I get, for example: [-2.4, 2.4, 1]
> 
> So, I guess I'm not doing things right, and the inequality constraints aren't being respected...
> From: Anders Peterson <an...@op...>
> To: oja...@li...
> Sent: Mon, June 13, 2011 4:55:38 PM
> Subject: Re: [ojAlgo-user] QuadraticSolver issue
> 
> What org.ojalgo.optimisation.State do you get?
> 
> Why don't you construct a junit test case to show me exactly what you're doing.
> 
> /Anders
> <TempClass.java><COVAR_MATRIX.txt><RETURN.txt><VARS.txt>------------------------------------------------------------------------------
> EditLive Enterprise is the world's most technically advanced content
> authoring tool. Experience the power of Track Changes, Inline Image
> Editing and ensure content is compliant with Accessibility Checking.
> http://p.sf.net/sfu/ephox-dev2dev_______________________________________________
> ojAlgo-user mailing list
> ojA...@li...
> https://lists.sourceforge.net/lists/listinfo/ojalgo-user