Approximate aggregation of Markovian models using alternating least squares
Performance Evaluation
| Hi-index | 0.00 |
Markovian models often compositional modeling andthe aggregation of components are used. Several approximateaggregation methods exist which are usually based onheuristics. This paper presents a new aggregation approachfor Markovian components which computes aggregates thatminimize the difference according to some algebraicallydefined function which describes the difference between thecomponent and the aggregate. If the difference becomes zero, aggregation is exact and component and aggregate are indistinguishable. Approximate aggregates are computed using analternating least squares approach which tries to minimizethe norm-wise difference between the original componentand the aggregate. The approach is extended to generatebounding aggregates which allow one to compute boundson transient or stationary quantities when the aggregate isembedded in an environment.