| dc.contributor.author | Rayas-Sánchez, José E. | |
| dc.contributor.author | Bandler, John W. | |
| dc.date.accessioned | 2013-05-21T15:23:19Z | |
| dc.date.available | 2013-05-21T15:23:19Z | |
| dc.date.issued | 2002-06 | |
| dc.identifier.citation | J. E. Rayas-Sánchez and J.W. Bandler, “EM-based design through neural space mapping methods,” in IEEE MTT-S Int. Microwave Symp. Workshop Notes and Short Courses, Seattle, WA, Jun. 2002. | es |
| dc.identifier.uri | http://hdl.handle.net/11117/585 | |
| dc.description | Artificial Neural Networks (ANN) and Space Mapping (SM) are eficiently combined to formulate EM-based design
algorithms. Neural Space Mapping (NSM) optimization and Neural Inverse Space Mapping (NISM) optimization are
reviewed.
NSM optimization exploits the SM-based neuromodeling techniques to efficiently approximate the mapping. The next
point is predicted avoiding parameter extraction (PE). The initial mapping is established by performing upfront fine model
analyses at a reduced number of base points. Coarse model sensitivities are exploited to select those base points. Huber
optimization is used to train, without testing points, simple SM-based neuromodels at each NSM iteration. EM-based yield
optimization is efficiently realized after NSM optimization.
NISM optimization is the first space mapping algorithm that explicitly makes use of the inverse of the mapping from the
fine to the coarse model parameter spaces. NISM follows an aggressive formulation by not requiring a number of up-front
fine model evaluations to start building the mapping. An statistical procedure to PE avoids the need for multipoint matching
and frequency mappings. It can also overcome poor local mimima during PE. An ANN whose generalization performance is
controlled through a network growing strategy approximates the inverse mapping at each iteration. In this manner, the ANN
always starts from a 2-layer perceptron and automatically migrates to a 3-layer perceptron only if the amount of nonlinearity
found in the inverse mapping becomes significant. The NISM step consists of evaluating the current neural network at the
optimal coarse solution. This step is equivalent to a quasi-Newton step while the inverse mapping is essentially linear, and
gradually departs from a quasi-Newton step as the amount of nonlinearity in the inverse mapping grows.
Contrast is made between neural space mapping design methods. A number of industrially relevant microwave design
problems are efficiently solved. | es |
| dc.description.sponsorship | ITESO, A.C. | es |
| dc.language.iso | eng | es |
| dc.publisher | IEEE MTT-S Int. Microwave Symp. Workshop Notes and Short Courses | es |
| dc.relation.ispartofseries | IEEE MTT-S Int. Microwave Symp. Workshop Notes and Short Courses;2002 | |
| dc.rights.uri | http://quijote.biblio.iteso.mx/licencias/CC-BY-NC-ND-2.5-MX.pdf | es |
| dc.subject | Neural Inverse Space Mapping | es |
| dc.subject | Microwave Circuits | es |
| dc.subject | Neural Networks | es |
| dc.subject | Space Mapping | es |
| dc.subject | Electromagnetic Based Design | es |
| dc.title | EM-Based Design through Neural Space Mapping Methods | es |
| dc.type | info:eu-repo/semantics/conferencePaper | es |
| rei.revisor | IEEE MTT-S Int. Microwave Symp. Workshop Notes and Short Courses | |
| rei.peerreviewed | No | es |
