intial notes.md

characteristics for successful algorithm:

Scalar Metric	Pass/Fail equivalent
how many of the constraints did it satisfy	did it satisfy all of the constraints?
how many feasible regions did it find	did it find all the feasible regions?
what was the variance of the distribution	was the variance above X?
was it able to produce enough points fast enough	does it make Y points-per-second?

another maybe good property: does it spit out localized failures?

some algorithms to test, as per George Cheng's recomendations in OPYL format:

Simple problem: Pressure vessel

inputs:
  Ts: {lower: 1.0, upper: 1.375}
  R: {lower: 25.0, upper: 150.0}
  L: {lower: 25.0, upper: 240.0}
  Th: {lower: 0.625, upper: 1.0}
  
constraints:
  c1: Ts - 0.0193*R >= 0
  c2: Th - 0.00954*R >= 0
  c3: (pi*(R^2)*L) + ((4/3)*pi*(R^3)) - (1.296*(10^6))>0

Medium problem: stepped-end beam

% lb=[0.01 0.3 0.5]; Replicated by n, number of beams
% ub=[0.05 0.65 1]; Replicated by n, number of beams

% n=10;
% E=2e11;
% P=50000;
% sigma=14000e4*2.5;
% for i=1:n
%     s1=0;
%     for j=1:i
%         s1=s1+x(3*j,:);
%     end
%     c(i,:)=6*P*s1./(x(3*i-2).*x(3*i-1).^2)-sigma;
%     c(i+n,:)=(x(3*i-1)./x(3*i-2))-25;
% end
% s2=0;
% for i=1:n
%     s2=s2+x(3*i,:).*x(3*i-1,:).*x(3*i-2,:);
% end
% c(2*n+1,:)=s2-0.4;
% %  c
% %  c1;
% %  c=[];
% ceq=[];

aka

constraints:

Hard problem: P118

% lb = [8;43;3;0;0;0;0;0;0;0;0;0;0;0;0];
% ub = [21;57;16;90;120;60;90;120;60;90;120;60;90;120;60];

% ccnt = 1;
% for i = 1:4
%     c(ccnt,:) = -x(3*i+1,:)+x(3*i-2,:)-7;
%     c(ccnt+1,:) = x(3*i+1,:)-x(3*i-2,:)-6;
%     c(ccnt+2,:) = -x(3*i+2,:)+x(3*i-1,:)-7;
%     c(ccnt+3,:) = x(3*i+2,:)-x(3*i-1,:)-7;
%     c(ccnt+4,:) = -x(3*i+3,:)+x(3*i,:)-7;
%     c(ccnt+5,:) = x(3*i+3,:)-x(3*i,:)-6;
%     ccnt = ccnt +6;
% end
% c(ccnt,:) = -x(1,:)-x(2,:)-x(3,:)+60;
% c(ccnt+1,:) = -x(4,:)-x(5,:)-x(6,:)+50;
% c(ccnt+2,:) = -x(7,:)-x(8,:)-x(9,:)+70;
% c(ccnt+3,:) = -x(10,:)-x(11,:)-x(12,:)+85;
% c(ccnt+4,:) = -x(13,:)-x(14,:)-x(15,:)+100;
% ceq = [];

For the matlab codes, dimension 1 (rows) represents variables/constraints. Dimension 2 (columns) represents points.