repeating elements of a vector

I have two vectors, say lambda=[1; 2] and ng=[3;4]. I want to have a long vector where the elements of lambda are repeated the corresponding times of the values of ng, i.e., I want to get

lambda_long=[1;1;1;2;2;2;2]

I can do it with some for loop, but is there any other efficient way of doing this?

Answer 1

[Oleg Komarov]

Fully vectorized run-length decoding:

% Example inputs (as vector columns)

lambda = [2; 4;9]; 
ng     = [3;4;7];

% Preallocate output

out  = zeros(sum(ng),1);

% Distribute starting point of sequences

csng = cumsum(ng);
out([1; csng(1:end-1)+1]) = [lambda(1); diff(lambda)];

% Propagate

out = cumsum(out)

Answer 2

[Andrei Bobrov]

x = arrayfun(@(a,b)a(ones(b,1)),lambda,ng,'un',0)
lambda_long = cat(1,x{:})

add

eg:

lambda = [14 5 89 47]'
ng=[3 4 2 4]'

% solution
id = zeros(sum(ng),1);
id(cumsum(ng)-ng+1)=1;
lambda_long = lambda(cumsum(id));

[Rabeya]

This is clever too. Thanks!

Answer 3

[Jan]

Similar to Oleg's encoding, but avoid DIFF/CUMSUM of data to prevent rounding errors:

lambda = [2; 4; 9]; 
ng     = [3; 4; 7];

index(cumsum(ng)+1) = 1;
index(1)   = 1;
index(end) = [];
out        = lambda(cumsum(index));

Untested!

Answer 4

[Wayne King]

One way

   lambda=[1; 2]; 
   ng=[3;4];
   A = repmat(lambda(1),ng(1),1);
   B = repmat(lambda(2),ng(2),1);
   C = [A;B];

Of course, you could put this on one line.

   C = [repmat(lambda(1),ng(1),1); repmat(lambda(2),ng(2),1)];

[Rabeya]

this is fine for such low length lambda, but not when you have 50 values for lambda!