Accuracy of Neural Network Approximators in Simulation-Optimization

Johnson, Virginia M.; Rogers, Leah L.

doi:10.1061/(asce)0733-9496(2000)126:2(48)

Cited by 110 publications

(53 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The proposed framework has roots in surrogate modeling which is concerned with developing and utilizing more efficient but lower-fidelity models as surrogates of computationally intensive highfidelity (original) models [Razavi et al, 2012]. There are two general families of surrogate modeling strategies: (1) response surface surrogates which utilize statistical or datadriven function approximation methods such as kriging and neural networks to emulate one (or multiple) response(s) of an original model [Broad et al, 2010;Johnson and Rogers, 2000;Regis and Shoemaker, 2007;Yan and Minsker, 2011], and (2) lower-fidelity physically based surrogates which are simplified models of the original system of interest and conceptually or directly preserve its governing physics [Forrester et al, 2007;Mondal et al, 2010;Robinson et al, 2008]. The proposed framework in this study shares some features with lower-fidelity physically based surrogate modeling when used for model calibration.…”

Section: Motivation and Objectivementioning

confidence: 99%

An efficient framework for hydrologic model calibration on long data periods

Razavi

Tolson

2013

Water Resour. Res.

View full text Add to dashboard Cite

.[1] Long periods of hydrologic data records have become available in many watersheds around the globe. Hydrologic model calibration on such long, full-length data periods is typically deemed the most robust approach for calibration but at larger computational costs. Determination of a representative short period as a ''surrogate'' of a long data period that sufficiently embeds its information content is not trivial and is a challenging research question. The representativeness of such a short period is not only a function of data characteristics but also model and calibration error function dependent. Unlike previous studies, this study goes beyond identifying the best surrogate data period to be used in model calibration and proposes an efficient framework that calibrates the hydrologic model to full-length data while running the model only on a short period for the majority of the candidate parameter sets. To this end, a mapping system is developed to approximate the model performance on the full-length data period based on the model performance for the short data period. The basic concepts and the promise of the framework are demonstrated through a computationally expensive hydrologic model case study. Three calibration approaches, namely calibration solely to a surrogate period, calibration to the full period, and calibration through the proposed framework, are evaluated and compared. Results show that within the same computational budget, the proposed framework leads to improved or equal calibration performance compared to the two conventional approaches. Results also indicate that model calibration solely to a short data period may lead to a range of performances from poor to very well depending on the representativeness of the short data period which is typically not known a priori.

show abstract

Section: Motivation and Objectivementioning

confidence: 99%

An efficient framework for hydrologic model calibration on long data periods

Razavi

Tolson

2013

Water Resour. Res.

View full text Add to dashboard Cite

show abstract

“…This is virtually reduced to near zero with ANN as the simulator. Johnson and Rogers (2000) have, concluded that ANN replaces the full model. In the present study, since the ANN is trained within the full range in which each decision variable is varied, it is indeed true that ANN fully replaces the SHARP simulation model.…”

Section: Computational Timementioning

confidence: 99%

Optimal Groundwater Management in Deltaic Regions using Simulated Annealing and Neural Networks

Rao¹,

Thandaveswara

Bhallamudi

et al. 2003

Water Resources Management

View full text Add to dashboard Cite

Abstract. This study deals with the optimal management of groundwater in deltaic aquifer systems with some reference to east coastal hydro-geo-climatic conditions of India. A system of cooperative wells is proposed to supplement surface water sources to meet the demand during the non-monsoon season, without inducing excessive saltwater intrusion. The management models are solved as nonlinear, non-convex, combinatorial problems. The management models are solved by interfacing simulated annealing (SA) algorithm with an existing SHARP interface flow model to determine an optimal policy for location and pumpages of cooperative wells. Computational burden arising from SA algorithm is managed within practical timeframes by replacing the simulator with an artificial neural network (ANN).

show abstract

“…(1) methods that reduce the execution time required for the simulation model through parallel algorithms and computer architectures (Dougherty 1991;Tompson et al 1994); and (2) methods that use an approximation of the simulation model, called a meta-model, to quickly supply predictions during the course of the search (Johnson & Rogers 2000). This latter approach is the focus of the current study.…”

Section: Introductionmentioning

confidence: 99%

“…ANNs impose fewer constraints on the functional form of the relationships between input and output variables, making them a logical choice for application when the complexity of the mapping is difficult to anticipate. Multilayer perceptrons trained by a backpropagation learning algorithm have been successfully used in modeling complex relations such as rainfall -runoff processes (Smith & Eli 1995), prediction of daily stream flows (Sureerattanan & Phein 1997), forecasting water quality parameters (Maier & Dandy 1996), inferring reservoir operating rules (Raman & Chanramoulia 1996;Ponnambalam et al 2003;Mousavi et al 2007), groundwater systems operation and conjunctive-use modeling (Ranjithan et al 1993;Rogers & Dowla 1994;Johnson & Rogers 2000). Broad et al (2005) …”

Section: Introductionmentioning

confidence: 99%