Are there any other way that I can optimize my functions or write them differently?

I'm trying to grab a values from a function with an input of time. With my code right now, it takes about 10 minutes to evaluate 100 data points, and 20 minutes for 180 data points. I believe what takes the function long to run is the Arterial function. Is there any other ways I can write this code? I'm new to Matlab.

The following are my functions:

This function grab a value from predictionvalue function.

function grabvalues= grabvalue(tfinal,F,f,z,PS)
  tic
  grabvalues=zeros(1,length(tfinal));
  for i=1:length(tfinal)
      grabvalues(i)= predictionvalue(tfinal(i),F,f,z,PS);
  end
  toc
  end

This function has three main parts; Arterial function, RK4 function, and Sampling time. The Arterial function takes in shortdata.txt file and linear interpolate the output. The RK4 then use the Arterial function to get the values for Sample times.
  function  tis_predicted2 = predictionvalue(tfinal,F,f,z,PS)
  %----------------Define Functions-----------%
  %Arterial Function
  load shortdata.txt
  tA=shortdata(:,1);
  AA=shortdata(:,2);
  function output=AF_function_in(tA_input)
  %size(tA);
  xq1=0:0.01:tA_input;
  vq1=interpn(tA,AA,xq1,'linear');
  output=vq1(end);
  %plot(xq1,vq1,'-',tA,AA,'o');
  %set(gca,'XLim',[0 10])
  end

    %Two Compartmental Model Functions
    fP=@(t,I,P) (F/f)*(AF_function_in(t)- P) - (PS/f)*(P-I); % Hct(mv)=0.45
    fI=@(t,I,P)  (PS/z)*(P - I);

    %Initial Condition
    h=0.01;
    N=ceil(tfinal/h);
    t=zeros(1,N);
    P=zeros(1,N);
    I=zeros(1,N);

    %------------------- RK4 Loop-----------------------------------%
    for i=1:N %i=[0, tfinal/h]
        % Update t
        t(i+1)=t(i)+h;
        %Update equation
        k1P = fP(t(i)      , I(i)           ,      P(i)    );
        k1I = fI(t(i)      , I(i)           ,      P(i)    );

        k2P = fP(t(i)+0.5*h, I(i)+ 0.5*k1I*h, P(i)+0.5*k1P*h);
        k2I = fI(t(i)+0.5*h, I(i)+ 0.5*k1I*h, P(i)+0.5*k1P*h);

        k3P = fP(t(i)+0.5*h, I(i)+ 0.5*k2I*h, P(i)+0.5*k2P*h);
        k3I = fI(t(i)+0.5*h, I(i)+ 0.5*k2I*h, P(i)+0.5*k2P*h);

        k4P = fP(t(i)+h    ,I(i)+k3I*h      , P(i)+ k3P*h);
        k4I = fI(t(i)+h    ,I(i)+k3I*h      , P(i)+ k3P*h);

        I(i+1) = (I(i) + h/6 *(k1I + 2*k2I + 2*k3I + k4I));
        P(i+1) = (P(i) + h/6 *(k1P + 2*k2P + 2*k3P + k4P));   
    end
    Tissue_predicted = P*f+I*z;

    %-----------------SAMPLING ---------------------------------------------%
    tdelta=0.835;
    tfinalsub=tfinal+0.000001;
    tSamp=zeros(1,ceil(tfinalsub/tdelta));
    Samp=zeros(1,ceil(tfinalsub/tdelta));
    a(1)=0;
    for d = 1:ceil(tfinalsub/tdelta)-1
        a(d+1)= a(d)+tdelta/h;
        tSamp(d+1)= tSamp(d)+tdelta;
        Samp(d+1)= Tissue_predicted(1,ceil(a(d)+1));   
    end 
    tis_predicted2 = Samp(end);
    end

Answer 1

[Steven Lord]

"I believe what takes the function long to run is the Arterial function."

I recommend that you profile your code and use that data to identify the sections that take the most time, to be certain.

Just at a glance, my educated guess about one of the bottlenecks in your code will be this line.

load shortdata.txt

Don't load the data once per iteration of the for loop in your grabvalue function. Load it once before that loop starts and pass the data into the predictionvalue function as an additional input.

Also, look at the Code Analyzer warnings, if any, in your code. Correcting any issues it identifies may improve the performance and/or robustness of your code.

[Arbol]

I'm trying to come up with a code that would load the data before the prediction function comes in. However, I couldn't come up with anything. My thought was trying to use a classdef that could call the data values and use method to do it?

[Steven Lord]

You're passing five inputs (tfinal(i), F, f, z, and PS) into predictionvalue when you call it. Why not pass a sixth containing the data you loaded?