Question

anyone here got matlab??

Answer 1

matrixlaboratory

Answer 2

The Language of Tec Computing?

Answer 3

botmean do u have matlab??

Answer 4

I don't man

Answer 5

please go download

Answer 6

I should

Answer 7

So is this a free advertisement for them hehe :P

Answer 8

no lol i need someone that can run mycodee

Answer 9

i do @dan815

Answer 10

really!! ok wait

Answer 11

can u see if this code runs 4 u

Answer 12

@dan815 its saying page not found

Answer 13

are you on your computer or phone

Answer 14

computer

Answer 15

ok in that case how about if u try copy pasting this code

Answer 16

function out=w03assignment2_method3()
%Generating Normal set of 1000 points, a random set of numbers added to
%each other over and over should yield a normally distributed set of data
%in this case the mean should be 0.5 if i divide by the total normal of
%data sets added together.


%Step by Step of what my code is doing:
% 1)First generate normally distributed data points to sample from
% 2)defined my piecewise function in 'mclf' function and generated plot

%3) determined a started plane pos, and plane error, at each increment the
%plane's position changed +100+/- error.
%4) Randomly selected 75 points, set all their weight to 100% or 1, then
%the plane's detected height was acquired by mclf(plane_pos), this was
%taken to be a guassian distribution at that point, wiht 500 std. This was
%used to determine the weight, of the random sampled points.
%w=p(xsamp)/p(current). This weight updated the current weight of the point
%by taking a mean between the 2.
% 5)If the weight dropped below some percentage. The point was dropped and,
% another point closer to the higher weighted points was taken into
% account.
% 6)I created the sigmadiff function and made use of the normally
% distributed data, to pick a normally distributed value and see the
% standard deviation away from the mean, this way I will get more
% deviations that are closer to the mean, and I devised a method where for
% every 0.5 standard deviation I picked a lower weight point to distribute
% the random point. This way not all the random points went to the highest
% weight point. However a larger number did as that point is weighted
% heigher to begin with.
% 7)the new weight determined the new size, and thats about it.
% 8)For this method there is something different going on when it comes to
% the resampling stage, since the rejection criteria is very high, as soon
% as the weight decreases below 0.9 it is discarded, however, the resampled
% points are put at the lowest weight and set to 1, which bumps them down
% the bottom of the list, when sorted befor each for loop, and in the next
%iteration if they were put beside a good point at the top of the list,
%they gradually cycle back up instead of dissappearing on the next check down, if
% not they quickly drop below the rejection criteria and are discarded and
% put beside another lower weighted point, but the lower weighted points
% are stabilizing at the top as a result of cycling up the whole matrix,
% so they are the converging points, and the new points sampled there are
% even better gusses or very close to the right gusses, with this method we keep
%producing points that are dropping in probability slower and slower, so you will be
%saturated with high probability points all throughout the list, all above 0.9 prob atleast.
% ------------------------
n_data=rand(1000,1)
for i=1:999
n_data=n_data+rand(1000,1);
end
n_data=n_data./1000
std=sum((n_data-mean(n_data)).^2)/sqrt(1000)
n_mean=mean(n_data)


x=1:10000
for j= 1:10000
piece(j)=mclf(j);
end

%axis([0 10000 0 9000])
%plot(x,y)

%planepos
p_error=20*(rand-0.5)
%plane starting at pos =0
p_pos=0

g=ones(75,2)
g(1:75,1)=rand(75,1)*10000

count=0
while p_pos<=10000
count=count+1
clf
p_error=20*(rand-0.5);
p_pos=p_pos+100+p_error;
hold all;
plot(p_pos,8000,'o','MarkerSize',20)
%plane localization(plane mean location)
plot((mean(g(:,1))-100),8900,'*','MarkerSize',15)

plot(x,piece)
axis([0 10111 0 9200])
%
%sorting the 75 by 2 matrix of position and weights with respect to the
%weights of the points
g=sortrows(g,2);
%redistribute the weighting, the highest weight is always 1, the most
%correlated weight gets bumped to a 1 and everythign is wrt to that.
%this code below and the comment above is for method 2
%g(:,2)=g(:,2)/max(g(:,2))

for k=1:75
plot(g(k,1),8500,'o','MarkerSize',g(k,2)*10)

%this gets the new weight from p(trial)/p(current) and then takes the mean
%with the existing weight at that point and replaced the weight with it.
g(k,2)=(g(k,2)+(gauss(mclf(g(k,1)),mclf(p_pos))/gauss(mclf(p_pos),mclf(p_pos))))/2;

%if the weight of the point is less than 90% or located past the 10,000
%mark we replace the point with another guess, i made my guess normally
%distributed around the highest weights with the help of the sigmadigg
%function i defined at the bottom, every 0.5 deviation i get from my mean
%from my normally distributed data, it will be one of m highest weighted
%points to place the new guess around.
if g(k,2)<0.90 | g(k,1)>10000
%numbo=randint(1,1,[1,1000]);
numbo=randi(1000);
guess=n_data(numbo);

%this is the main difference in method 3, i am throwing my random guesses
%around the worst weight which ended up there as a result of slowly cycling
%up, the bad points dissappear faster without slowly cycling up to the
%lowest weight
g(k,1)=g((sigmadiff(n_mean,guess,std)),1)+200*(rand-0.5);
g(k,2)=1
%g(k,2)=g((sigmadiff(n_mean,guess,std)),2);
end


%
% if g(k,1)>=10000
% guess=n_data(1,1,randint([1,1000]))
% g(k,1)=g(1,sigmadiff(n_mean,guess,std))+20*(rand-0.5)
% g(k,2)=g(2,sigmadiff(n_mean,guess,std))
%




% if g(k,1)>=10000
% g(k,1)=((count-1)*100+500-(count-1)*100)*rand+(count-1)*100
% end
% if g(k,2)<0.05
% g(k,1)=((count-1)*100+500-(count-1)*100)*rand+(count-1)*100
% g(k,2)=0.9
% end
%
end

g(:,1)=g(:,1)+100
%
%g(1:75,1)=g(1:75,1)+100 ;

%
% plot(g,8500,'o','MarkerSize',10)

p=getframe;

end
end

%-----FuNCTION DEFINITIONs
function out=mclf(x)
if 1<= x & x <=1000
out=0;
elseif 1000<x & x<=2000
out=x-1000;
elseif 2000<x & x<=3000
out=x.*0.5;
elseif 3000<x & x<=4000
out=x.*0.25+750;
elseif 4000<x & x<=5000
out=x*0.6-650;
elseif 5000<x & x<=6000
out=x.*-0.4+4350
elseif 6000<x & x<=7000
out=1950;
elseif 7000<x & x<=8000
out=x.*-0.6+6150;
elseif 8000<x & x<=9000
out=x.*(-0.9)+8550
elseif 9000<x & x<=10000
out=450;
elseif x>10000
out=0;
end
end

%Gaussian distribution function
function out =gauss(x,m)
sig=500;
out=1/(sig*sqrt(2*pi)) * exp(-1/2*(((8000-x)-(8000-m))/sig)^2);
end

%picewise def function
function out=mclf(x)
if 1<= x & x <=1000
out=0;
elseif 1000<x & x<=2000
out=x-1000;
elseif 2000<x & x<=3000
out=x.*0.5;
elseif 3000<x & x<=4000
out=x.*0.25+750;
elseif 4000<x & x<=5000
out=x*0.6-650;
elseif 5000<x & x<=6000
out=x.*-0.4+4350
elseif 6000<x & x<=7000
out=1950;
elseif 7000<x & x<=8000
out=x.*-0.6+6150;
elseif 8000<x & x<=9000
out=x.*(-0.9)+8550
elseif 9000<x & x<=10000
out=450;
elseif x>10000
out=450;
end
end

%sigmadiff func
function out=sigmadiff(m,x,std)
t=abs((m-x)/std);
if 0<=t<0.5
out=1;
elseif 0.5<=t<1
out=2;
elseif 1<=t<1.5
out=3;
elseif 1.5<=t<2
out=4;
elseif 2<=t<2.5
out=5;
elseif 2.5<=t<3
out=6;
elseif 3<=t
out=1;
end
end

Answer 17

this page may help

http://www.mathworks.com/help/matlab/matlab_prog/nested-functions.html

it explains how to nest functions

Answer 18

oh and you also have two identical lines that state "function out=mclf(x)"

when I deleted one copy, the function worked for me

Answer 19

did u get the error for the downloaded code or the copy pasted not, which one i gotta fix its running fine for me

Answer 20

the thing is it might be running fine for me beccause i alread have files with the functions at the bottom defined so i wanna see if it runs for someone else who doesnt have the functiosn defined on their computer but now put at the bottom of the code

Answer 21

it runs fine for me I think, I get this after the animation plays out

Answer 22

ohh !! thanks for catching that double function hehe

Answer 23

here is my new code see how it works, basically the * is the plane approximation and the Big Circle is the true position of the plane and the small circles are the approximations for the plane location and the size of the small circles are how probable it is there

Answer 24

the first one is supposed to be the fast version and the 2nd one pauses to show how the new sample points are being regenerated

Answer 25

could you tell me if the 2nd one is running like way too slow on your computer

Answer 26

it runs pretty fast for me but my comp is fast, not sure how fast itll run on the school comps

Answer 27

the point of the simulation is to see how fast you can get the algorithm to localize aroung the point

Answer 28

when I try to run the second one, I get

"Undefined function or variable 'w03assignment2_method3_f2'."

Answer 29

oh right, forgot to match up the names

Answer 30

the second one is definitely different (not sure how to explain what's going on)

Answer 31

the 2nd one is yeah, im not sure how to explain it aswell i just think its like showing the new sampled points after ever single change