Types Of Model Structures Research Articles

Fast estimation of multiple group generalized linear latent variable models for categorical observed variables

A computationally efficient method for marginal maximum likelihood estimation of multiple group generalized linear latent variable models for categorical data is introduced. The approach utilizes second-order Laplace approximations of the integrals in the likelihood function. It is demonstrated how second-order Laplace approximations can be utilized highly efficiently for generalized linear latent variable models by considering symmetries that exist for many types of model structures. In a simulation with binary observed variables and four correlated latent variables in four groups, the method has similar bias and mean squared error compared to adaptive Gauss-Hermite quadrature with five quadrature points while substantially improving computational efficiency. An empirical example from a large-scale educational assessment illustrates the accuracy and computational efficiency of the method when compared against adaptive Gauss-Hermite quadrature with three, five, and 13 quadrature points.

Deep clustering of cooperative multi-agent reinforcement learning to optimize multi chiller HVAC systems for smart buildings energy management

Chillers are responsible for almost half of the total energy demand in buildings. Hence, the obligation of control systems of multi-chiller due to changes indoor environments is one of the most significant parts of a smart building. Such a controller is described as a nonlinear and multi-objective algorithm, and its fabrication is crucial to achieving the optimal balance between indoor thermal comfort and running a minimum number of chillers. This work proposes deep clustering of cooperative multi-agent reinforcement learning (DCCMARL) as well-suited to such system control, which supports centralized control by learning of agents. In MARL, since the learning of agents is based on discrete sets of actions and stats, this drawback significantly affects the model of agents for representing their actions with efficient performance. This drawback becomes considerably worse when increasing the number of agents, due to the increased complexity of solving MARL, which makes modeling policy very challenging. Therefore, the DCCMARL of multi-objective reinforcement learning is leveraging powerful frameworks of a hybrid clustering algorithm to deal with complexity and uncertainty, which is a critical factor that influences to the achievement of high levels of a performance action. The results showed that the ability of agents to manipulate the behavior of the smart building could improve indoor thermal conditions, as well as save energy up to 44.5% compared to conventional methods. It seems reasonable to conclude that agents' performance is influenced by what type of model structure.

Shaking table test on 1000 kV UHV transmission tower-line coupling system

Ultra high voltage (UHV) transmission tower-line coupling system consists of transmission lines (conductors and ground wires) supported by a series of lattice towers. However, in the current seismic design code, the transmission lines are just simplified as additional mass while the effect of nonlinear vibration of transmission lines on the seismic responses of transmission towers is not taken into consideration. To analyze the influence of the ignorance of the coupling effect between towers and transmission lines, two types of model structures, i.e. tower-line coupling system model and single tower model with lumped masses, should be tested for comparison. But normal reduced-scale model design method cannot complete the shaking table test by a uniform geometric similarity ratio for the coupling system in the limit of table size, hence this paper proposed to rescale the geometric similarity ratio for the model design of transmission lines and meanwhile satisfied the requirement in dynamic properties similitude and inertia force similitude separately, which successfully solved the problem that the span of transmission lines in the test was too large to fit within the shaking table. This method may offer reference to experimental studies for other large-span structures and the experimental results show that the nonlinear vibration of transmission lines plays a role as seismic reduction by dissipating some seismic energy to towers under earthquake.

Robust Product Line Design

The majority of approaches to product line design that have been proposed by marketing scientists assume that the underlying choice model that describes how the customer population will respond to a new product line is known precisely. In reality, however, marketers do not precisely know how the customer population will respond and can only obtain an estimate of the choice model from limited conjoint data. In this paper, we propose a new type of optimization approach for product line design under uncertainty. Our approach is based on the paradigm of robust optimization where, rather than optimizing the expected revenue with respect to a single model, one optimizes the worst-case expected revenue with respect to an uncertainty set of models. This framework allows us to account for parameter uncertainty, when we may be confident about the type of model structure but not about the values of the parameters, and structural uncertainty, when we may not even be confident about the right model structure to use to describe the customer population. Through computational experiments with a real conjoint data set, we demonstrate the benefits of our approach in addressing parameter and structural uncertainty. With regard to parameter uncertainty, we show that product lines designed without accounting for parameter uncertainty are fragile and can experience worst-case revenue losses as high as 23%, and that the robust product line can significantly outperform the nominal product line in the worst case, with relative improvements of up to 14%. With regard to structural uncertainty, we similarly show that product lines that are designed for a single model structure can be highly suboptimal under other structures (worst-case losses of up to 37%), while a product line that optimizes against the worst of a set of structurally distinct models can outperform single model product lines by as much as 55% in the worst case and can guarantee good aggregate performance over structurally distinct models.

Model Structure Choice for a Static VAR Compensator Under Modeling Uncertainty and Incomplete Information

To simulate the complex behavior of power systems, operators frequently rely on models. The task of model identification and validation becomes important in this context. The validity of the models has a direct influence on operator’s decisions and actions. In other words, erroneous or imprecise models lead to erroneous predictions of the systems’ behavior which may result in unwanted operator’s actions. This paper addresses the challenge of model structure choice for modeling and parameter identification in power systems. Three types of model structures are analyzed: 1) physical principle-based modeling; 2) black-box modeling (NARX, transfer function, Hammerstein–Wiener model); and 3) combination of physical and black-box modeling. This analysis has been performed using real grid measurements and available knowledge about a static VAR compensator (SVC) connected to the U.K.’s transmission network and operated by National Grid. The SVC’s modeling is presented in the context of a generalized modeling and identification algorithm, that is offered as a guideline for engineers. The model validity issues of the identified SVC models that include modeling uncertainty are discussed.

Modeling inoculum dose dependent patterns of acute virus infections

Inoculum dose, i.e. the number of pathogens at the beginning of an infection, often affects key aspects of pathogen and immune response dynamics. These in turn determine clinically relevant outcomes, such as morbidity and mortality. Despite the general recognition that inoculum dose is an important component of infection outcomes, we currently do not understand its impact in much detail. This study is intended to start filling this knowledge gap by analyzing inoculum dependent patterns of viral load dynamics in acute infections. Using experimental data for adenovirus and infectious bronchitis virus infections as examples, we demonstrate inoculum dose dependent patterns of virus dynamics. We analyze the data with the help of mathematical models to investigate what mechanisms can reproduce the patterns observed in experimental data. We find that models including components of both the innate and adaptive immune response are needed to reproduce the patterns found in the data. We further analyze which types of innate or adaptive immune response models agree with observed data. One interesting finding is that only models for the adaptive immune response that contain growth terms partially independent of viral load can properly reproduce observed patterns. This agrees with the idea that an antigen-independent, programmed response is part of the adaptive response. Our analysis provides useful insights into the types of model structures that are required to properly reproduce observed virus dynamics for varying inoculum doses. We suggest that such models should be taken as basis for future models of acute viral infections.

Sparse Estimation of Polynomial and Rational Dynamical Models

In many practical situations, it is highly desirable to estimate an accurate mathematical model of a real system using as few parameters as possible. At the same time, the need for an accurate description of the system behavior without knowing its complete dynamical structure often leads to model parameterizations describing a rich set of possible hypotheses; an unavoidable choice, which suggests sparsity of the desired parameter estimate. An elegant way to impose this expectation of sparsity is to estimate the parameters by penalizing the criterion with the $\ell_{0}$ “norm” of the parameters. Due to the non-convex nature of the $\ell_{0}$ -norm, this penalization is often implemented as solving an optimization program based on a convex relaxation (e.g., $\ell_{1}$ /LASSO, nuclear norm, $\ldots$ ). Two difficulties arise when trying to apply these methods: (1) the need to use cross-validation or some related technique for choosing the values of regularization parameters associated with the $\ell_{1}$ penalty; and (2) the requirement that the (unpenalized) cost function must be convex. To address the first issue, we propose a new technique for sparse linear regression called SPARSEVA, with close ties with the LASSO (least absolute shrinkage and selection operator), which provides an automatic tuning of the amount of regularization. The second difficulty, which imposes a severe constraint on the types of model structures or estimation methods on which the $\ell_{1}$ relaxation can be applied, is addressed by combining SPARSEVA and the Steiglitz-McBride method. To demonstrate the advantages of the proposed approach, a solid theoretical analysis and an extensive simulation study are provided.

Sensitivity analysis for seismic response of reinforced concrete staggered wall structures

This study investigates the sensitivity of design variables such as ultimate strength of concrete, yield stress of reinforcing steel and damping ratio to the seismic response of staggered wall structures. Two different methods for sensitivity analysis, the tornado diagram method and the first-order second-moment method, are applied to two types of model structures. Non-linear dynamic analyses are carried out using the seven maximum considered earthquake level earthquake records, which shows that the inter-storey drift is the most sensitive to the variation of modal damping ratio. It is observed in the incremental dynamic analyses that, when the earthquake intensity is relatively small, the yield stress of rebars and the concrete strength in the link beams are also important factors affecting the sensitivity of seismic response. As the intensity of seismic load increases, the strength of columns becomes another important factor.

Seismic Performance of Building Structures with Spatial Variability

In this study the seismic performances of reinforced concrete apartment buildings with Y-shaped plans having central core are investigated. Three types of model structures are designed for each shape depending on the amount of shear partition walls: structures with all shear walls, structures with all columns except the core walls, and structures with shear walls and columns combined. The required amount of concrete is the largest in the all shear wall structures, and decreases as more and more shear walls are replaced with columns. According to nonlinear static and dynamic analysis results, the structures with all shear walls and all columns turn out to have the largest and the smallest strengths, respectively. Index Terms—Seismic performance evaluation, fragility analysis.

Sparse Estimation of Rational Dynamical Models

In many practical situations, it is highly desirable to estimate an accurate mathematical model of a real system using as few parameters as possible. This can be motivated either from appealing to a parsimony principle (Occam's razor) or from the view point of the utilization complexity in terms of control synthesis, prediction, etc. At the same time, the need for an accurate description of the system behavior without knowing its complete dynamical structure often leads to model parameterizations describing a rich set of possible hypotheses; an unavoidable choice, which suggests sparsity of the desired parameter estimate. An elegant way to impose this expectation of sparsity is to estimate the parameters by penalizing the criterion with the l0 norm of the parameters, which is often implemented as solving an optimization program based on a convex relaxation (e.g. l1/LASSO, nuclear norm, …). However, in order to apply these methods, the (unpenalized) cost function must be convex. This imposes a severe constraint on the types of model structures or estimation methods on which these relaxations can be applied. In this paper, we extend the use of convex relaxation techniques for sparsity to general rational plant model structures estimated by using prediction error minimization. This is done by combining the LASSO and the Steiglitz-McBride approaches. To demonstrate the advantages of the proposed solution an extensive simulation study is provided.

Identification of Nonlinear Feedback Systems Using a Structured Polynomial Nonlinear State Space Model

In this paper, we propose an identification procedure for nonlinear feedback systems by introducing a new type of model structure. The idea is to combine the advantages of two different model structures: the Polynomial NonLinear State Space (PNLSS) model and the block-structured feedback model. By imposing structural information to the existing PNLSS model, we obtain a flexible model with a lower number of parameters compared with the unconstrained structure. The structured-PNLSS model consists of a linear block in the feedforward path and a PNLSS block in the feedback path. The presented identification procedure for the structured-PNLSS model is applied to input/output measurements of a nonlinear device.

Two-mode clustering methods: astructuredoverview

In this paper we present a structured overview of methods for two-mode clustering, that is, methods that provide a simultaneous clustering of the rows and columns of a rectangular data matrix. Key structuring principles include the nature of row, column and data clusters and the type of model structure or associated loss function. We illustrate with analyses of symptom data on archetypal psychiatric patients.

An automated relational database structure building system for the life cycle of manufacturing simulation models

Recent simulation literature shows a growing interest in the integration of artificial intelligent concepts with computer simulation to automate the simulation phases of model construction and translation. Little attention has been paid to aiding modellers in the use of output data over the life cycle of a simulation model associated with a manufacturing system study. In modern computer integrated manufacturing (CIM) environments, information from related areas ideally should cohabit a common database system. In this paper, relational database structures for computer simulation results over the life cycle of simulation models are developed by classifying different types of model structures, user goals, and associated experimental design conditions common to all manufacturing systems. The structures provide an intermediate step between a pure simulation environment and a complete CIM relational database system and serve as a formal basis to create an automated database structure building system. Several simulation environments have been constructed with this approach and have proven successful.

Summary of the process identification workshop at the 1992 Canadian chemical engineering conference

Abstract The highlights of the workshop can be summarized as follows: (1) The workshop reaffirmed that data prefiltering is a necessary step in the model building process in order to ensure nonstationary disturbances are not modelled as part of the process; (2) In addition, the workshop confirmed that improvements in the steady-state gain are seen if input test signals are carefully designed using the disturbance dynamics and the expected movement of the manipulated variables while under control as guides. As a result, the majority of input test signals should have significantly more power at low frequencies in order to improve the steady-state gain estimates; (3) Data prefiltering and input signal design decisions appear to affect the final model quality much more than the type of model structure used (Box-Jenkins, finite impulse response, frequency sampling filter); (4) Although the newer techniques presented by some of the university groups do not necessarily yield better quality models, they simplify the model building process by automating model building selections, or by providing better descriptions of model inadequacies.

Knowledge-based Generation Of The AssemblyOrder Of Elements In Wooden Construction

In order to develop new, improved systems for building construction, it is necessary to evaluate the various factors related to construction methodology. This paper presents a method for generating the assembly order of building elements in wooden housing structures based on an information about the characteristics of the building construction systems such as physical considerations about the positions of building elements relative to one another, or the jointing method employed, etc. In order to derive criteria for judging the assembly order of wooden elements, 2,355 connections in three common types of building construction systems for wooden houses currently in use in Japan were surveyed. For each of these connections, data related to the assembly order and its determinants were examined. A portion of the information was drawn from the results of building site surveys conducted within the last four years, and specialists in the field were interviewed as well. Based on the results of the survey, 16 separate judging rules for assembly order were deduced. In these rules, knowledge about the relationship between each of the discrete elements was obtained through inference based on characteristic data of the building construction systems. The knowledge-based system (KBS) employed by the author to generate the assembly sequence composed of inference engine, utility module, knowledge base about assembly rules and judgement rules on the relation between two elements. The system was constructed by using Prolog, whereas the input data was composed of list-structured data. Through the application of this system to some different types of model structures, it is proved that we can get the proper assembly order of all structural elements. Therefore, the system can be used as a tool for evaluating newly developed wooden building construction systems from the point of view of construction steps.

Cotorsion pairs and degreewise homological model structures

Let C be an abelian category. We show that under certain hypotheses, a cotorsion pair (A,B) in C may induce two natural homological model structures on Ch(C). One is such that the (trivially) cofibrant objects form the class of (exact) complexes A for which each An 2 A. The other is such that the (trivially) fibrant objects form the class of (exact) complexes B for which each Bn 2 B. Special cases of these model structures such as Hovey’s “locally free” model structure and “flasque” model structure have already appeared in the literature. The examples support the belief that any useful homological model structure comes from a single cotorsion pair on the ground category C. Furthermore, one of the two types of model structures we consider requires surprisingly few assumptions to exist. For example, Theorem 4.7 implies that every cotorsion pair (A,B) of R-modules which is cogenerated by a set gives rise to a cofibrantly generated homological model structure on Ch(R).