Behavioural queueing : cellular automata and a laboratory experiment

Dans cette thèse, nous étudions les aspects comportementaux d'agents qui interagissent dans des systèmes de files d'attente à l'aide de modèles de simulation et de méthodologies expérimentales. Chaque période les clients doivent choisir un prestataire de servivce. L'objectif est d'analyser l'impact des décisions des clients et des prestataires sur la formation des files d'attente. Dans un premier cas nous considérons des clients ayant un certain degré d'aversion au risque. Sur la base de leur perception de l'attente moyenne et de la variabilité de cette attente, ils forment une estimation de la limite supérieure de l'attente chez chacun des prestataires. Chaque période, ils choisissent le prestataire pour lequel cette estimation est la plus basse. Nos résultats indiquent qu'il n'y a pas de relation monotone entre le degré d'aversion au risque et la performance globale. En effet, une population de clients ayant un degré d'aversion au risque intermédiaire encoure généralement une attente moyenne plus élevée qu'une population d'agents indifférents au risque ou très averses au risque. Ensuite, nous incorporons les décisions des prestataires en leur permettant d'ajuster leur capacité de service sur la base de leur perception de la fréquence moyenne d'arrivées. Les résultats montrent que le comportement des clients et les décisions des prestataires présentent une forte "dépendance au sentier". En outre, nous montrons que les décisions des prestataires font converger l'attente moyenne pondérée vers l'attente de référence du marché. Finalement, une expérience de laboratoire dans laquelle des sujets jouent le rôle de prestataire de service nous a permis de conclure que les délais d'installation et de démantèlement de capacité affectent de manière significative la performance et les décisions des sujets. En particulier, les décisions du prestataire, sont influencées par ses commandes en carnet, sa capacité de service actuellement disponible et les décisions d'ajustement de capacité qu'il a prises, mais pas encore implémentées. - Queuing is a fact of life that we witness daily. We all have had the experience of waiting in line for some reason and we also know that it is an annoying situation. As the adage says "time is money"; this is perhaps the best way of stating what queuing problems mean for customers. Human beings are not very tolerant, but they are even less so when having to wait in line for service. Banks, roads, post offices and restaurants are just some examples where people must wait for service. Studies of queuing phenomena have typically addressed the optimisation of performance measures (e.g. average waiting time, queue length and server utilisation rates) and the analysis of equilibrium solutions. The individual behaviour of the agents involved in queueing systems and their decision making process have received little attention. Although this work has been useful to improve the efficiency of many queueing systems, or to design new processes in social and physical systems, it has only provided us with a limited ability to explain the behaviour observed in many real queues. In this dissertation we differ from this traditional research by analysing how the agents involved in the system make decisions instead of focusing on optimising performance measures or analysing an equilibrium solution. This dissertation builds on and extends the framework proposed by van Ackere and Larsen (2004) and van Ackere et al. (2010). We focus on studying behavioural aspects in queueing systems and incorporate this still underdeveloped framework into the operations management field. In the first chapter of this thesis we provide a general introduction to the area, as well as an overview of the results. In Chapters 2 and 3, we use Cellular Automata (CA) to model service systems where captive interacting customers must decide each period which facility to join for service. They base this decision on their expectations of sojourn times. Each period, customers use new information (their most recent experience and that of their best performing neighbour) to form expectations of sojourn time at the different facilities. Customers update their expectations using an adaptive expectations process to combine their memory and their new information. We label "conservative" those customers who give more weight to their memory than to the xiv Summary new information. In contrast, when they give more weight to new information, we call them "reactive". In Chapter 2, we consider customers with different degree of risk-aversion who take into account uncertainty. They choose which facility to join based on an estimated upper-bound of the sojourn time which they compute using their perceptions of the average sojourn time and the level of uncertainty. We assume the same exogenous service capacity for all facilities, which remains constant throughout. We first analyse the collective behaviour generated by the customers' decisions. We show that the system achieves low weighted average sojourn times when the collective behaviour results in neighbourhoods of customers loyal to a facility and the customers are approximately equally split among all facilities. The lowest weighted average sojourn time is achieved when exactly the same number of customers patronises each facility, implying that they do not wish to switch facility. In this case, the system has achieved the Nash equilibrium. We show that there is a non-monotonic relationship between the degree of risk-aversion and system performance. Customers with an intermediate degree of riskaversion typically achieve higher sojourn times; in particular they rarely achieve the Nash equilibrium. Risk-neutral customers have the highest probability of achieving the Nash Equilibrium. Chapter 3 considers a service system similar to the previous one but with risk-neutral customers, and relaxes the assumption of exogenous service rates. In this sense, we model a queueing system with endogenous service rates by enabling managers to adjust the service capacity of the facilities. We assume that managers do so based on their perceptions of the arrival rates and use the same principle of adaptive expectations to model these perceptions. We consider service systems in which the managers' decisions take time to be implemented. Managers are characterised by a profile which is determined by the speed at which they update their perceptions, the speed at which they take decisions, and how coherent they are when accounting for their previous decisions still to be implemented when taking their next decision. We find that the managers' decisions exhibit a strong path-dependence: owing to the initial conditions of the model, the facilities of managers with identical profiles can evolve completely differently. In some cases the system becomes "locked-in" into a monopoly or duopoly situation. The competition between managers causes the weighted average sojourn time of the system to converge to the exogenous benchmark value which they use to estimate their desired capacity. Concerning the managers' profile, we found that the more conservative Summary xv a manager is regarding new information, the larger the market share his facility achieves. Additionally, the faster he takes decisions, the higher the probability that he achieves a monopoly position. In Chapter 4 we consider a one-server queueing system with non-captive customers. We carry out an experiment aimed at analysing the way human subjects, taking on the role of the manager, take decisions in a laboratory regarding the capacity of a service facility. We adapt the model proposed by van Ackere et al (2010). This model relaxes the assumption of a captive market and allows current customers to decide whether or not to use the facility. Additionally the facility also has potential customers who currently do not patronise it, but might consider doing so in the future. We identify three groups of subjects whose decisions cause similar behavioural patterns. These groups are labelled: gradual investors, lumpy investors, and random investor. Using an autocorrelation analysis of the subjects' decisions, we illustrate that these decisions are positively correlated to the decisions taken one period early. Subsequently we formulate a heuristic to model the decision rule considered by subjects in the laboratory. We found that this decision rule fits very well for those subjects who gradually adjust capacity, but it does not capture the behaviour of the subjects of the other two groups. In Chapter 5 we summarise the results and provide suggestions for further work. Our main contribution is the use of simulation and experimental methodologies to explain the collective behaviour generated by customers' and managers' decisions in queueing systems as well as the analysis of the individual behaviour of these agents. In this way, we differ from the typical literature related to queueing systems which focuses on optimising performance measures and the analysis of equilibrium solutions. Our work can be seen as a first step towards understanding the interaction between customer behaviour and the capacity adjustment process in queueing systems. This framework is still in its early stages and accordingly there is a large potential for further work that spans several research topics. Interesting extensions to this work include incorporating other characteristics of queueing systems which affect the customers' experience (e.g. balking, reneging and jockeying); providing customers and managers with additional information to take their decisions (e.g. service price, quality, customers' profile); analysing different decision rules and studying other characteristics which determine the profile of customers and managers.

Gene Ontology Function prediction in Mollicutes using Protein-Protein Association Networks

Abstract Background: Many complex systems can be represented and analysed as networks. The recent availability of large-scale datasets, has made it possible to elucidate some of the organisational principles and rules that govern their function, robustness and evolution. However, one of the main limitations in using protein-protein interactions for function prediction is the availability of interaction data, especially for Mollicutes. If we could harness predicted interactions, such as those from a Protein-Protein Association Networks (PPAN), combining several protein-protein network function-inference methods with semantic similarity calculations, the use of protein-protein interactions for functional inference in this species would become more potentially useful. Results: In this work we show that using PPAN data combined with other approximations, such as functional module detection, orthology exploitation methods and Gene Ontology (GO)-based information measures helps to predict protein function in Mycoplasma genitalium. Conclusions: To our knowledge, the proposed method is the first that combines functional module detection among species, exploiting an orthology procedure and using information theory-based GO semantic similarity in PPAN of the Mycoplasma species. The results of an evaluation show a higher recall than previously reported methods that focused on only one organism network.

Evolution of Coordination in Social Networks: A Numerical Study

Coordination games are important to explain efficient and desirable social behavior. Here we study these games by extensive numerical simulation on networked social structures using an evolutionary approach. We show that local network effects may promote selection of efficient equilibria in both pure and general coordination games and may explain social polarization. These results are put into perspective with respect to known theoretical results. The main insight we obtain is that clustering, and especially community structure in social networks has a positive role in promoting socially efficient outcomes.

Learning about Bayesian networks for forensic interpretation : An example based on the "the problem of multiple propositions"

Both, Bayesian networks and probabilistic evaluation are gaining more and more widespread use within many professional branches, including forensic science. Notwithstanding, they constitute subtle topics with definitional details that require careful study. While many sophisticated developments of probabilistic approaches to evaluation of forensic findings may readily be found in published literature, there remains a gap with respect to writings that focus on foundational aspects and on how these may be acquired by interested scientists new to these topics. This paper takes this as a starting point to report on the learning about Bayesian networks for likelihood ratio based, probabilistic inference procedures in a class of master students in forensic science. The presentation uses an example that relies on a casework scenario drawn from published literature, involving a questioned signature. A complicating aspect of that case study - proposed to students in a teaching scenario - is due to the need of considering multiple competing propositions, which is an outset that may not readily be approached within a likelihood ratio based framework without drawing attention to some additional technical details. Using generic Bayesian networks fragments from existing literature on the topic, course participants were able to track the probabilistic underpinnings of the proposed scenario correctly both in terms of likelihood ratios and of posterior probabilities. In addition, further study of the example by students allowed them to derive an alternative Bayesian network structure with a computational output that is equivalent to existing probabilistic solutions. This practical experience underlines the potential of Bayesian networks to support and clarify foundational principles of probabilistic procedures for forensic evaluation.

A Robust Multiple Feature Approach To Endpoint Detection In Car Environment Based On Advanced Classifiers

In this paper we propose an endpoint detection system based on the use of several features extracted from each speech frame, followed by a robust classifier (i.e Adaboost and Bagging of decision trees, and a multilayer perceptron) and a finite state automata (FSA). We present results for four different classifiers. The FSA module consisted of a 4-state decision logic that filtered false alarms and false positives. We compare the use of four different classifiers in this task. The look ahead of the method that we propose was of 7 frames, which are the number of frames that maximized the accuracy of the system. The system was tested with real signals recorded inside a car, with signal to noise ratio that ranged from 6 dB to 30dB. Finally we present experimental results demonstrating that the system yields robust endpoint detection.

EEG-based functional networks in schizophrenia.

Schizophrenia is often considered as a dysconnection syndrome in which, abnormal interactions between large-scale functional brain networks result in cognitive and perceptual deficits. In this article we apply the graph theoretic measures to brain functional networks based on the resting EEGs of fourteen schizophrenic patients in comparison with those of fourteen matched control subjects. The networks were extracted from common-average-referenced EEG time-series through partial and unpartial cross-correlation methods. Unpartial correlation detects functional connectivity based on direct and/or indirect links, while partial correlation allows one to ignore indirect links. We quantified the network properties with the graph metrics, including mall-worldness, vulnerability, modularity, assortativity, and synchronizability. The schizophrenic patients showed method-specific and frequency-specific changes especially pronounced for modularity, assortativity, and synchronizability measures. However, the differences between schizophrenia patients and normal controls in terms of graph theory metrics were stronger for the unpartial correlation method.

An allele-specific, stochastic gene expression process controls the expression of multiple Ly49 family genes and generates a diverse, MHC-specific NK cell receptor repertoire.

Mouse NK cells express MHC class I-specific inhibitory Ly49 receptors. Since these receptors display distinct ligand specificities and are clonally distributed, their expression generates a diverse NK cell receptor repertoire specific for MHC class I molecules. We have previously found that the Dd (or Dk)-specific Ly49A receptor is usually expressed from a single allele. However, a small fraction of short-term NK cell clones expressed both Ly49A alleles, suggesting that the two Ly49A alleles are independently and randomly expressed. Here we show that the genes for two additional Ly49 receptors (Ly49C and Ly49G2) are also expressed in a (predominantly) mono-allelic fashion. Since single NK cells can co-express multiple Ly49 receptors, we also investigated whether mono-allelic expression from within the tightly linked Ly49 gene cluster is coordinate or independent. Our clonal analysis suggests that the expression of alleles of distinct Ly49 genes is not coordinate. Thus Ly49 alleles are apparently independently and randomly chosen for stable expression, a process that directly restricts the number of Ly49 receptors expressed per single NK cell. We propose that the Ly49 receptor repertoire specific for MHC class I is generated by an allele-specific, stochastic gene expression process that acts on the entire Ly49 gene cluster.

Cognate microRNA-Gene Regulatory Networks Mediating Glioblastoma Oncogenic Properties

Malignant gliomas, including the most common and fatal form glioblastoma (GBM, WHO grade IV astrocytoma), remain a challenge to treat. In the United States and Europe, more than 30,000 patients per year are newly diagnosed with GBM. Despite ongoing trials, the best currently available multimodal treatment approaches include surgical resection followed by concomitant and adjuvant radiation (RT) and temozolomide (TMZ) therapy, resulting in a low median overall survival (OS) rate ranging from 12.2 - 15.9 months. The important role of genetic and epigenetic changes in DNA, RNA, and protein alteration as well as epigenetic changes secondary to the tumor microenvironment and outside selection pressure (therapeutic interventions), are increasingly being recognized. In GBM treatment, the focus is shifting toward a more patient-centered (personalized) therapy. In this regard, in particular, microRNAs are being increasingly studied. MicroRNAs are non¬protein coding small RNAs that serve as negative gene regulators by binding to a specific sequence in the promoter region of a target gene, thus regulating gene expression. A single microRNA potentially targets hundreds of genes; thus, microRNAs and their cognate target genes have important roles as tumor suppressors and oncogenes as well as regulators of various cancer- specific cellular features, such as proliferation, apoptosis, invasion, and metastasis. The identification of distinct microRNA-gene regulatory networks in GBM patients can be expected to provide novel therapeutic insights by identifying candidate patients for targeted therapies. To this end, in this work we identified and validated clinically relevant and meaningful novel gene- microRNA regulatory networks that correlated with MR tumor phenotypes, histopathology, and patient survival and response rates to therapy. - Le traitement des gliomes malins, y compris sous leur forme la plus commune et meurtrière, le glioblastome (GBM, ou astrocytome de grade IV selon l'OMS), demeure à ce jour un défi. Aux États-Unis et en Europe, un nouveau diagnostic de GBM est prononcé dans plus de 30Ό00 cas par an. En dépit de tests en cours, les meilleures approches thérapeutiques combinées actuellement disponibles comprennent la résection chirurgicale de la tumeur, suivie d'une radiothérapie adjuvante ainsi que d'un traitement au temozolomide (RT/TMZ), thérapies dont résulte une médiane de survie globale basse (overall survival, OS), comprise entre 12.2 et 15.9 mois. On reconnaît de plus en plus le rôle majeur de l'ADN, de l'ARN et de l'altération des protéines ainsi que des modifications épigénétiques, secondaires par rapport au microenvironnement de la tumeur et à la pression de sélection extérieure (les interventions thérapeutiques). Dans le traitement du GBM, le centre d'intérêt se déplace vers une thérapie centrée sur le cas individuel du patient. Dans ce but, en particulier les microARN sont de plus en plus analysés. Les microARN sont de petits ARN non-codants (les protéines) qui servent de régulateurs négatifs de gènes en s'attachant à une séquence spécifique dans la région promotrice d'un gène-cible, régulant ainsi l'expression du gène. Un seul microARN cible potentiellement des centaines de gènes; on a ainsi découvert que les microARN et leurs gènes-cibles apparentés ont une fonction importante en tant que suppresseurs de tumeurs et d'oncogènes, ainsi que comme régulateurs de diverses caractéristiques cellulaires spécifiques du cancer, comme la prolifération, l'apoptose, l'invasion et la métastase. On peut s'attendre à ce que l'identification de réseaux microARN régulateurs de gènes, distincts selon les patients de GBM, fournisse une approche thérapeutique inédite par la détermination des patients susceptibles de réagir favorablement à des thérapies ciblées.

Stochastic resonance in a dipole

We show that the dipole, a system usually proposed to model relaxation phenomena, exhibits a maximum in the signal-to-noise ratio at a nonzero noise level, thus indicating the appearance of stochastic resonance. The phenomenon occurs in two different situations, i.e., when the minimum of the potential of the dipole remains fixed in time and when it switches periodically between two equilibrium points. We have also found that the signal-to-noise ratio has a maximum for a certain value of the amplitude of the oscillating field.

Statistical mechanics of multi-edge networks

Statistical properties of binary complex networks are well understood and recently many attempts have been made to extend this knowledge to weighted ones. There are, however, subtle yet important considerations to be made regarding the nature of the weights used in this generalization. Weights can be either continuous or discrete magnitudes, and in the latter case, they can additionally have undistinguishable or distinguishable nature. This fact has not been addressed in the literature insofar and has deep implications on the network statistics. In this work we face this problem introducing multiedge networks as graphs where multiple (distinguishable) connections between nodes are considered. We develop a statistical mechanics framework where it is possible to get information about the most relevant observables given a large spectrum of linear and nonlinear constraints including those depending both on the number of multiedges per link and their binary projection. The latter case is particularly interesting as we show that binary projections can be understood from multiedge processes. The implications of these results are important as many real-agent-based problems mapped onto graphs require this treatment for a proper characterization of their collective behavior.

Wireless Sensor Networks for Infrastructure Monitoring, TR-611, 2010

A good system of preventive bridge maintenance enhances the ability of engineers to manage and monitor bridge conditions, and take proper action at the right time. Traditionally infrastructure inspection is performed via infrequent periodical visual inspection in the field. Wireless sensor technology provides an alternative cost-effective approach for constant monitoring of infrastructures. Scientific data-acquisition systems make reliable structural measurements, even in inaccessible and harsh environments by using wireless sensors. With advances in sensor technology and availability of low cost integrated circuits, a wireless monitoring sensor network has been considered to be the new generation technology for structural health monitoring. The main goal of this project was to implement a wireless sensor network for monitoring the behavior and integrity of highway bridges. At the core of the system is a low-cost, low power wireless strain sensor node whose hardware design is optimized for structural monitoring applications. The key components of the systems are the control unit, sensors, software and communication capability. The extensive information developed for each of these areas has been used to design the system. The performance and reliability of the proposed wireless monitoring system is validated on a 34 feet span composite beam in slab bridge in Black Hawk County, Iowa. The micro strain data is successfully extracted from output-only response collected by the wireless monitoring system. The energy efficiency of the system was investigated to estimate the battery lifetime of the wireless sensor nodes. This report also documents system design, the method used for data acquisition, and system validation and field testing. Recommendations on further implementation of wireless sensor networks for long term monitoring are provided.

Stochastic inversion of tracer test and electrical geophysical data to estimate hydraulic conductivities.

Quantifying the spatial configuration of hydraulic conductivity (K) in heterogeneous geological environments is essential for accurate predictions of contaminant transport, but is difficult because of the inherent limitations in resolution and coverage associated with traditional hydrological measurements. To address this issue, we consider crosshole and surface-based electrical resistivity geophysical measurements, collected in time during a saline tracer experiment. We use a Bayesian Markov-chain-Monte-Carlo (McMC) methodology to jointly invert the dynamic resistivity data, together with borehole tracer concentration data, to generate multiple posterior realizations of K that are consistent with all available information. We do this within a coupled inversion framework, whereby the geophysical and hydrological forward models are linked through an uncertain relationship between electrical resistivity and concentration. To minimize computational expense, a facies-based subsurface parameterization is developed. The Bayesian-McMC methodology allows us to explore the potential benefits of including the geophysical data into the inverse problem by examining their effect on our ability to identify fast flowpaths in the subsurface, and their impact on hydrological prediction uncertainty. Using a complex, geostatistically generated, two-dimensional numerical example representative of a fluvial environment, we demonstrate that flow model calibration is improved and prediction error is decreased when the electrical resistivity data are included. The worth of the geophysical data is found to be greatest for long spatial correlation lengths of subsurface heterogeneity with respect to wellbore separation, where flow and transport are largely controlled by highly connected flowpaths.