I have a set of N documents, each of some maximum size M. Each document is of different length and contains a series of word features of size K. The end of the document is padded with zeros(K, 1) features up to maximum size. Therefore, my data is in a matrix of size (K, N*M).
For each document I compute some number H based on document size, giving me a vector of size (N, 1) of all the Hs. I only want to get the data for each document up to index H and then concatenate all those features together. So let's say N = 3, M = 100, K = 10, data, which is of dimensionality (10, 3*100), and H = (45, 30, 25) I would want to have result = [data(1:45) data(101:131) data(201:226)]. Is there a way to do this without a for loop? I want to make the code efficient.
edit: Sorry, I realized there is an easier way to ask my question. Say I have a long vector, data, a vector of start indices, st, and a vector of end indices, ed. I want to get data([st(1):ed(1) st(2):ed(2) st(3):ed(3)…]). Is there a way to do this?