Arne Drud

CONOPT: A GRG code for large sparse dynamic nonlinear optimization problems

The paper presents CONOPT, an optimization system for static and dynamic large-scale nonlinearly constrained optimization problems. The system is based on the GRG algorithm. All computations involving the Jacobian of the constraints use sparse-matrix algorithms from linear programming, modified to deal with the nonlinearity and to take maximum advantage of the periodic structure in dynamic models. The paper presents the main features of the system, espcially the inversion routines and their data structures, the dynamic setting of tolerances in Newton’s algorithm, and the user features in the overal packaging. The difficulties with implementing a practical GRG algorithm are described in detail. Computational experience with some medium to large models is presented, idicating the viability of CONOPT for certain real-life problems, particularly those involving almost as many constraints as variables.

MINLPLib--A Collection of Test Models for Mixed-Integer Nonlinear Programming

The paper describes a new computerized collection of test models for mixed-integer nonlinear programming. Because there is no standard format for nonlinear models, the model collection is augmented with a translation server that can transform the models from their basic GAMS format into other formats, including AMPL, BARON, LGO, LINGO, and MINOPT. The translation server can also be used to transform industrial models that contain confidential information. Such transformations allow many of these models to be distributed to the research community as highly relevant algorithmic test models.

Macroeconomic modeling based on social-accounting principles

Abstract Using the example of a small comparative static model of Thailand for 1980, this article sets out an approach to macroeconomic model building that is based on two versions of a social-accounting matrix (SAM)—one version contains data for a base year, whereas the cell entries of the other are algebraic expressions for the determination of the corresponding transaction values. Thus the model is developed in transaction-value (TV) form, and this is one distinguishing feature of the approach. Other features derive from different aspects of the relationship between the two SAMs. We argue that the approach has distinct advantages for model description, calibration, and solution and that these are important if models are to be used for policy purposes that place a premium on intelligibility and replicability within the context of a flexible modeling capability.

Balancing large social accounting matrices with nonlinear network programming

The Social Accounting Matrix (SAM) framework is a convenient way of giving a comprehensive and consistent picture of an economy. A SAM database can only be used as the basis of a planning model if it is balanced, in the sense that it satisfies a set of internal consistency relations. This paper describes a nonlinear network optimization model for balancing the SAM framework in order that it will conform to a set of internal consistency relations.

A survey of model representations and simulation algorithms in some existing modeling systems

Abstract The purpose of the paper is to explore the possibilities for automatically communicating models between modeling systems. The main difficulty in model communication seems to be the difference in information requirements between solution algorithms. The paper therefore discusses the difference between a model and a solution algorithm, and it describes the most common simulation algorithms from the perspective of information requirements. It continues to survey sixteen econometric modeling systems, concentrating on how equations and simulation models are defined, and on the correlation between model formats and simulation algorithms. Other aspects covered are if-then-else constructions, more general model types like optimal control models, and indexing.

Interactions between nonlinear programming and modeling systems

Modeling systems are very important for bringing mathematical programming software to nonexpert users, but few nonlinear programming algorithms are today linked to a modeling system. The paper discussed the advantages of linking modeling systems with nonlinear programming. Traditional algorithms can be linked using black-box function and derivatives evaluation routines for local optimization. Methods for generating this information are discussed. More sophisticated algorithms can get access to almost any type of information: interval evaluations and constraint restructuring for detailed preprocessing, second order information for sequential quadratic programming and interior point methods, and monotonicity and convex relaxations for global optimization. Some of the sophisticated information is available today; the rest can be generated on demand.

Macroeconomic adjustment in Thailand: Demand management and supply conditions

This 1983 article, Macroeconomic Adjustment in Thailand: Demand Management and Supply Conditions, was reprinted with permission from the Journal of Policy Modeling. Within the framework of a fix-price, four-sector (agriculture, industry, energy, and services) macroeconomic model for Thailand, comparative statics are used to assess alternative ways of macroeconomic adjustment. The methods discussed are fiscal policy interventions, manipulations of the exchange rate, and productivity improvements. Fiscal policies include changing the relative prices between domestically produced, imported, and exported goods, and increased protection via higher import tariffs or subsidization of exports. Productivity improvements through structural changes include changes in the conditions of agricultural production, the use of energy, the productivity of primary factors of production, and the conditions governing import behavior. The article derives the implications of these adjustment methods in terms of income generation, external deficit, and inflation. It is shown that only productivity improvements have positive effects on all indicators. Fiscal interventions lead to an improvement in the external deficit, but at the cost of income generation. The outcome of devaluation is largely dependent on the behavior of factor prices.

Quality Assurance and Global Optimization

GAMS Development and ARKI Consulting use Quality Assurance (QA) as an integral part of their software development and software publishing process. Research and development in the global optimization area has resulted in promising implementations. Initiated by customer demand, we have been adding three different global codes, BARON, LGO, and OQNLP, to our portfolio of nonlinear optimization solvers. As part of our QA effort towards the integration and testing of these new global solvers an open architecture for reliability and performance testing of mixed-integer nonlinear optimization codes has been released. This open testing framework has been placed in the public domain (www.gamsworld.org) to serve our customers, and researchers in general, by making reproducible tests a practical proposition. We give examples illustrating the quality assurance process for obtaining performance results for local and global nonlinear and mixed-integer nonlinear programming solvers, using the existing framework tools described in this article.

Hercules : a modeling system with knowledge about economics

The paper is concerned with the effectiveness of economic modeling. It argues that the key problems of economic modeling are the time and effort it takes to implement a model, the problems of detecting and correcting errors, and the difficulties involved in communicating assumptions. The logical consequence is to improve the model representation and the error detection capabilities of modeling systems. The paper discusses information content and error detection in models that use conventional mathematical notation and argues that we need a model representation with more information than is usually available in mathematical representations. HERCULES, a modeling systems for large economywide static models is then presented. The system uses a model representation in which all economic assumptions are directly available. This makes model implementation much faster and prevents certain types of errors. In other cases, error detection and error messages are far superior to those of general purpose mathematical modeling systems. The technical implementation of HERCULES, designed to allow maximum flexibility for future enhancements is discussed.

Interfacing modeling systems and solution algorithms

Abstract The paper discusses the demands an algorithm developer has to a modeling system when he adds a new algorithm to the system, and it discusses how algorithms should be designed so they later on can be added to several independent modeling systems. Two practical examples are used to make the issues clearer. The purpose of the paper is to make both developers of modeling systems and of algorithms aware of each others needs and problems, such that future cooperation can be improved.