Search for: All records

Design researchers have struggled to produce quantitative predictions for exactly why and when diversity might help or hinder design search efforts. This paper addresses that problem by studying one ubiquitously used search strategy—Bayesian optimization (BO)—on a 2D test problem with modifiable convexity and difficulty. Specifically, we test how providing diverse versus non-diverse initial samples to BO affects its performance during search and introduce a fast ranked-determinantal point process method for computing diverse sets, which we need to detect sets of highly diverse or non-diverse initial samples. We initially found, to our surprise, that diversity did not appear to affect BO, neither helping nor hurting the optimizer’s convergence. However, follow-on experiments illuminated a key trade-off. Non-diverse initial samples hastened posterior convergence for the underlying model hyper-parameters—a model building advantage. In contrast, diverse initial samples accelerated exploring the function itself—a space exploration advantage. Both advantages help BO, but in different ways, and the initial sample diversity directly modulates how BO trades those advantages. Indeed, we show that fixing the BO hyper-parameters removes the model building advantage, causing diverse initial samples to always outperform models trained with non-diverse samples. These findings shed light on why, at least for BO-type optimizers, the use of diversity has mixed effects and cautions against the ubiquitous use of space-filling initializations in BO. To the extent that humans use explore-exploit search strategies similar to BO, our results provide a testable conjecture for why and when diversity may affect human-subject or design team experiments.

 

Abstract Engineering systems are often composed of many subsystems that interact with each other. These subsystems, referred to as disciplines, contain many types of uncertainty and in many cases are feedback-coupled with each other. In designing these complex systems, one needs to assess the stationary behavior of these systems for the sake of stability and reliability. This requires the system level uncertainty analysis of the multidisciplinary systems, which is often computationally intractable. To overcome this issue, techniques have been developed for capturing the stationary behavior of the coupled multidisciplinary systems through available data of individual disciplines. The accuracy and convergence of the existing techniques depend on a large amount of data from all disciplines, which are not available in many practical problems. Toward this, we have developed an adaptive methodology that adds the minimum possible number of samples from individual disciplines to achieve an accurate and reliable uncertainty propagation in coupled multidisciplinary systems. The proposed method models each discipline function via Gaussian process (GP) regression to derive a closed-form policy. This policy sequentially selects a new sample point that results in the highest uncertainty reduction over the distribution of the coupling design variables. The effectiveness of the proposed method is demonstrated in the uncertainty analysis of an aerostructural system and a coupled numerical example.