Clustering algorithms using stochastic analysis and ensemble techniques.
block_sum(a, blocks, axis=0)
Given an array and a list of index blocks along a given axis, delivers a new array whose indices correspond to blocks and whose entries are sums over the old values.
For example, consider the following matrix with the blocks indicated by horizontal lines, and the resulting sum over blocks:
\[\left[ \begin{array}{c|cc|c} 0 & 1 & 2 & 3\\\hline 4 & 5 & 6 & 7\\ 8 & 9 & 10 & 11 \end{array} \right] \mapsto \left[ \begin{array}{ccc} 0 & 3 & 3 \\ 12 & 30 & 18 \end{array} \right]\]This can be achieved by calling block_sum
twice,
first with blocks = [[0],[1,2]]
and axis=0
,
second with blocks = [[0],[1,2],[3]]
and axis=1
.