Scaling positive and negative weights

Hi, I have a vector that contains positive as well as negative weights. For example:

0.75
-0.14
-0.92
0.56
-0.28
...

Now I need a way to transform these weights so that...

• all weights will be positive.
• each weight will be greater than 0 and smaller than 1.
• the weights sum to 1.
• the more negative the original weight, the lower its new value should be.

How would I do that?

Answer 1

[Image Analyst]

Just scale linearly - at least that's one of an unlimited number of ways:

% Declare random sample data.
signal = 20*rand(1,10)-10
% Offset so that the min of the new signal is at zero.
newSignal = signal - min(signal);
% Make the new signal sum to 1
newSignal = newSignal / sum(newSignal)
% Print out the sum, just to check.
fprintf('The sum is %f.\n', sum(newSignal));

[Image Analyst]

I hope that wasn't your homework!!! I don't want to make you a plagiarist.

[dave]

Thanks, there's only the problem that one of weights will be zero. But I need all of them to be > 0. btw: Don't worry, this has nothing to do with any kind of homework..

[Image Analyst]

Add a very small amount, like eps

newSignal = signal - min(signal) + eps;