Aftab - If you want to slow down epsilon's approach to zero as k increases, then you could do something like the following
% create an array of linearly spaced numbers
epsilonVector = fliplr(linspace(eps,1,10000));
So the linspace function will create a vector of 10000 numbers that are linearly spaced between eps and 1. We then flip the vector from right to left so that epsilonVector(1) is 1, and epsilonVector(10000) is eps. If you want different lower and upper bounds on epsilon, then you can change the inputs to linspace.
I'm not sure how you want to use epsilon in your algorithm; perhaps an alternative approach is needed if you don't want linearly spaced points.
