|
DOUG 0.2
|
The algorithm is defined in SpMtx_aggregation::SpMtx_aggregate
, aggregate size bounds
,
, aggregation radius r and number of smoothing steps
non-overlapping) aggregates

for which
;
;
; 

; choose a seednode
from
(or randomly from set
if
)
;
and 


, add to
all
that are connected through
to at least 2 nodes in
, set
and set
.
(i.e. the largest layer) and add to
all
of shortest path length from 

that is too small (i.e.
) with a connected neighbouring aggregate
(subject to the requirement
; it may be necessary to split up
to achieve this) and shrink
accordingly.The smoothing is then done to get restriction matrix and coarse problem (see Smoothed Coarse Spaces).
1.7.3-20110217