In the BFR algorithm, since we assume all clusters are “axis…
In the BFR algorithm, since we assume all clusters are “axis-aligned” so that the covariance of a cluster can be constructed from the variables SUM, SUMSQ, N, which only requires memory of d + d + 1. However, if we drop this “axis-aligned” assumption, how much memory (in units) would be required for a cluster to compute the co-variance?
Read Details