An official website of the United States government
Machine learning system design frequently necessitates balancing multiple objectives, such as prediction error and energy consumption, for deep neural networks (DNNs). Typically, no single design performs well across all objectives; thus, finding Pareto-optimal designs is of interest. Measuring different objectives frequently incurs different costs; for example, measuring the prediction error of DNNs is significantly more expensive than measuring the energy consumption of a pre-trained DNN because it requires re-training the DNN. Current state-of-the-art methods do not account for this difference in objective evaluation cost, potentially wasting costly evaluations of objective functions for little information gain. To address this issue, we propose a novel cost-aware decoupled approach that weights the improvement of the hypervolume of the Pareto region by the measurement cost of each objective. To evaluate our approach, we perform experiments on several machine learning systems deployed on energy constraints environments.
more »
« less
Is Verifying Frequently an Optimal Strategy? A Belief-Based Model of Verification
Verification activities increase an engineering team’s confidence in its system design meeting system requirements, which in turn are derived from stakeholder needs. Conventional wisdom suggests that the system design should be verified frequently to minimize the cost of rework as the system design matures. However, this strategy is based more on experience of engineers than on a theoretical foundation. In this paper, we develop a belief-based model of verification of system design, using a single system requirement as an abstraction, to determine the conditions under which it is cost effective for an organization to verify frequently. We study the model for a broad set of growth rates in verification setup and rework costs. Our results show that verifying a system design frequently is not always an optimal verification strategy. Instead, it is only an optimal strategy when the costs of reworking a faulty design increase at a certain rate as the design matures.
more »
« less
- Award ID(s):
- 1762883
- PAR ID:
- 10201584
- Date Published:
- Journal Name:
- ASME 2020 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Machine learning system design frequently necessitates balancing multiple objectives, such as prediction error and energy consumption for deep neural networks (DNNs). Typically, no single design performs well across all objectives; thus, finding Pareto-optimal designs is of interest. Measuring different objectives frequently incurs different costs; for example, measuring the prediction error of DNNs is significantly more expensive than measuring the energy consumption of a pre-trained DNN because it requires re-training the DNN. Current state-of-the-art methods do not account for this difference in objective evaluation cost, potentially wasting costly evaluations of objective functions for little information gain. To address this issue, we propose a novel cost-aware decoupled approach that weights the improvement of the hypervolume of the Pareto region by the measurement cost of each objective. We perform experiments on a of range of DNN applications for comprehensive evaluation of our approach.more » « less
-
An optimal control problem is considered for a stochastic differential equation containing a state-dependent regime switching, with a recursive cost functional. Due to the non-exponential discounting in the cost functional, the problem is time-inconsistent in general. Therefore, instead of finding a global optimal control (which is not possible), we look for a time-consistent (approximately) locally optimal equilibrium strategy. Such a strategy can be represented through the solution to a system of partial differential equations, called an equilibrium Hamilton–Jacob–Bellman (HJB) equation which is constructed via a sequence of multi-person differential games. A verification theorem is proved and, under proper conditions, the well-posedness of the equilibrium HJB equation is established as well.more » « less
-
null (Ed.)Since passenger demand in urban transit systems is asymmetrically distributed across different periods in a day and different geographic locations across the cities, the tradeoff between vehicle operating costs and service quality has been a persistent problem in transit operational design. The emerging modular vehicle technology offers us a new perspective to solve this problem. Based on this concept, we propose a variable-capacity operation approach with modular transits for shared-use corridors, in which both dispatch headway and vehicle capacity are decision variables. This problem is rigorously formulated as a mixed integer linear programming model that aims to minimize the overall system cost, including passenger waiting time costs and vehicle operating costs. Because the proposed model is linear, the state-of-the-art commercial solvers (e.g., Gurobi) can be used to obtain the optimal solution of the investigated problem. With numerical experiments, we demonstrate the feasibility of the mathematical model, verify the effectiveness of the proposed model in reducing overall system costs in transit systems, as well as the robustness of the proposed model with different parameter settings.more » « less
-
Stathopoulos, Ted (Ed.)Vortex-induced vibration (VIV) of deck girders frequently drives the wind-resistant design of wind-sensitive bridges from the preliminary to final design stages. Shaping bridge decks is a proven strategy to mitigate VIV. While significant shape modifications are commonly restricted to preliminary design stages, only minor medications are possible at advanced design stages, typically involving adding flow modifiers or changing the shape and location of existing appendages. These mitigation strategies have been implemented in the last decades by carrying out expensive wind tunnel campaigns and following heuristic design rules. This paper proposes an experimental data-driven adaptive surrogate-based optimization approach to systematically identify optimum deck shapes that minimizes the economic cost of the bridge while fulfilling the VIV project specifications. The methodology is conceived to carry out simultaneously the general and detailed shape design of the final deck configuration by harnessing a sequential sampling plan aiming at reducing the sectional model construction costs. The proposed holistic design framework is successfully applied to a real application case involving 26 wind tunnel tests of 1.8-meter wide sectional models to figure out the optimal gap distance and location of maintenance tracks of a twin-box deck equipped with all the appendages included in the final deck design.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/doi.org/10.1115/DETC2020-22582
