Automatic model type selection with heterogeneous evolution: An application to RF circuit block modeling

Abstract

Many complex, real world phenomena are difficult to study directly using controlled experiments. Instead, the use of computer simulations has become commonplace as a cost effective alternative. However, regardless of Moore's law, performing high fidelity simulations still requires a great investment of time and money. Surrogate modeling (metamodeling) has become indispensable as an alternative solution for relieving this burden. Many surrogate model types exist (Support Vector Machines, Kriging, RBF models, Neural Networks, ...) but no type is optimal in all circumstances. Nor is there any hard theory available that can help make this choice. The same is true for setting the surrogate model parameters (BiasVariance trade-off). Traditionally, the solution to both problems has been a pragmatic one, guided by intuition, prior experience or simply available software packages. In this paper we present a more founded approach to these problems. We describe an adaptive surrogate modeling environment, driven by speciated evolution, to automatically determine the optimal model type and complexity. Its utility and performance is presented on a case study from electronics.

Particle Swarm Model Selection for Authorship Verification

“…PSMS is a more ambitious formulation that selects full models for classification without requiring much supervision [4]. Only a few methods have been proposed for facing the full model selection problem, most notably the work by Gorissen et al [5]. Unlike the latter method, PSMS is more efficient and simple to implement, moreover, PSMS has shown to be robust against overfitting because of the way the search is guided.…”

Surrogate Modeling of RF Circuit Blocks

“…In order to reduce the computational cost of such comparison, transistor level simulations have been replaced with an analytical model of the LNA 3 [4]. Although the accuracy of such analytical model is obviously not sufficient to replace the circuit simulator in the design process, it satisfactorily reproduces the shape of the simulator outputs.…”

“…Several surrogate model types have been compared (as implemented in the SUMO Toolbox): artificial neural networks, rational functions, radial basis functions, least squares support vector machines and kriging [4], [5].…”

Evolutionary Regression Modeling with Active Learning: An Application to Rainfall Runoff Modeling

“…A full description of the algorithm, model types, and genetic operators is out of scope for this paper. Such settings can be found in [5]. Given the correlation between the outputs, they are not modeled separately (by separate models) but together in a single model with multiple outputs.…”

“…See [4] for a good discussion on this issue. For the application in this paper we will utilize a fully featured toolbox for adaptive surrogate model generation , the SUMO toolbox [5].…”