Model Selection Under Limited Information Using a Value-of-Information-Based Indicator

Abstract

Designers are continuously challenged by complexity, as well as by the excessive instantiation and execution times of models, particularly in the context of integrated product and materials design. In order to manage these challenges, a systematic strategy for evaluating and selecting models is presented in this paper. The systematic strategy is based on value-of-information for design decision making. It consists of a (i) process performance indicator (PPI) to quantify the impact of model refinement from a decision-centric perspective and (ii) a method involving model evaluation. Using this method, a least complex but valid model is evaluated, and, only if necessary, gradually refined it until the most appropriate one is selected. The systematic approach is particularly well suited for integrated product and materials design, and all other scenarios where the perfect knowledge of the true system behavior and bounds of error are not available throughout the design space. The proposed strategy is applied to the design of photonic crystal waveguides for use in a next-generation optoelectronic communication system. In this paper, it is shown that the systematic strategy based on the PPI is useful for evaluating and selecting models particularly when accuracy of the prediction or the associated error bounds are not known.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 17 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.