Investigações sobre técnicas de arquivamento para otimizadores multiobjetivo

Multi-objective problems may have many optimal solutions, which together form the Pareto optimal set. A class of heuristic algorithms for those problems, in this work called optimizers, produces approximations of this optimal set. The approximation set kept by the optmizer may be limited or unlimited. The benefit of using an unlimited archive is to guarantee that all the nondominated solutions generated in the process will be saved. However, due to the large number of solutions that can be generated, to keep an archive and compare frequently new solutions to the stored ones may demand a high computational cost. The alternative is to use a limited archive. The problem that emerges from this situation is the need of discarding nondominated solutions when the archive is full. Some techniques were proposed to handle this problem, but investigations show that none of them can surely prevent the deterioration of the archives. This work investigates a technique to be used together with the previously proposed ideas in the literature to deal with limited archives. The technique consists on keeping discarded solutions in a secondary archive, and periodically recycle these solutions, bringing them back to the optimization. Three methods of recycling are presented. In order to verify if these ideas are capable to improve the archive content during the optimization, they were implemented together with other techniques from the literature. An computational experiment with NSGA-II, SPEA2, PAES, MOEA/D and NSGA-III algorithms, applied to many classes of problems is presented. The potential and the difficulties of the proposed techniques are evaluated based on statistical tests.

A methodology for hierarchical image segmentation evaluation

This paper proposes a method to evaluate hierarchical image segmentation procedures, in order to enable comparisons between different hierarchical algorithms and of these with other (non-hierarchical) segmentation techniques (as well as with edge detectors) to be made. The proposed method builds up on the edge-based segmentation evaluation approach by considering a set of reference human segmentations as a sample drawn from the population of different levels of detail that may be used in segmenting an image. Our main point is that, since a hierarchical sequence of segmentations approximates such population, those segmentations in the sequence that best capture each human segmentation level of detail should provide the basis for the evaluation of the hierarchical sequence as a whole. A small computational experiment is carried out to show the feasibility of our approach.

Flow Phenomena: Fluids, Heat and Mass

In this book are published results of high-tech application of computational modeling and simulation the dynamics of different flows, heat and mass transfer in different fields of science and engineering.

A computational study of some rheological influences on the ""splashing experiment""

In various attempts to relate the behaviour of highly-elastic liquids in complex flows to their rheometrical behaviour, obvious candidates for study have been the variation of shear viscosity with shear rate, the two normal stress differences N(1) and N(2), especially N(1), the extensional viscosity, and the dynamic moduli G` and G ``. In this paper, we shall confine attention to `constant-viscosity` Boger fluids, and, accordingly, we shall limit attention to N(1), eta(E), G` and G ``. We shall concentrate on the ""splashing"" problem (particularly that which arises when a liquid drop falls onto the free surface of the same liquid). Modern numerical techniques are employed to provide the theoretical predictions. We show that high eta(E) can certainly reduce the height of the so-called Worthington jet, thus confirming earlier suggestions, but other rheometrical influences (steady and transient) can also have a role to play and the overall picture may not be as clear as it was once envisaged. We argue that this is due in the main to the fact that splashing is a manifestly unsteady flow. To confirm this proposition, we obtain numerical simulations for the linear Jeffreys model. (C) 2010 Elsevier B.V. All rights reserved.

Changing musical emotion : a computational rule system for modifying score and performance

When communicating emotion in music, composers and performers encode their expressive intentions through the control of basic musical features such as: pitch, loudness, timbre, mode, and articulation. The extent to which emotion can be controlled through the systematic manipulation of these features has not been fully examined. In this paper we present CMERS, a Computational Music Emotion Rule System for the control of perceived musical emotion that modifies features at the levels of score and performance in real-time. CMERS performance was evaluated in two rounds of perceptual testing. In experiment I, 20 participants continuously rated the perceived emotion of 15 music samples generated by CMERS. Three music works, each with five emotional variations were used (normal, happy, sad, angry, and tender). The intended emotion by CMERS was correctly identified 78% of the time, with significant shifts in valence and arousal also recorded, regardless of the works’ original emotion.

Correlating flow induced shear stress and chondrocytes activity in micro-porous scaffold using computational fluid dynamics and rapid prototyping

Flow induced shear stress plays an important role in regulating cell growth and distribution in scaffolds. This study sought to correlate wall shear stress and chondrocytes activity for engineering design of micro-porous osteochondral grafts based on the hypothesis that it is possible to capture and discriminate between the transmitted force and cell response at the inner irregularities. Unlike common tissue engineering therapies with perfusion bioreactors in which flow-mediated stress is the controlling parameter, this work assigned the associated stress as a function of porosity to influence in vitro proliferation of chondrocytes. D-optimality criterion was used to accommodate three pore characteristics for appraisal in a mixed level fractional design of experiment (DOE); namely, pore size (4 levels), distribution pattern (2 levels) and density (3 levels). Micro-porous scaffolds (n=12) were fabricated according to the DOE using rapid prototyping of an acrylic-based bio-photopolymer. Computational fluid dynamics (CFD) models were created correspondingly and used on an idealized boundary condition with a Newtonian fluid domain to simulate the dynamic microenvironment inside the pores. In vitro condition was reproduced for the 3D printed constructs seeded by high pellet densities of human chondrocytes and cultured for 72 hours. The results showed that cell proliferation was significantly different in the constructs (p<0.05). Inlet fluid velocity of 3×10-2mms-1 and average shear stress of 5.65×10-2 Pa corresponded with increased cell proliferation for scaffolds with smaller pores in hexagonal pattern and lower densities. Although the analytical solution of a Poiseuille flow inside the pores was found insufficient for the description of the flow profile probably due to the outside flow induced turbulence, it showed that the shear stress would increase with cell growth and decrease with pore size. This correlation demonstrated the basis for determining the relation between the induced stress and chondrocyte activity to optimize microfabrication of engineered cartilaginous constructs.

A comparison of the gas phase acidities of phospholipid headgroups: Experimental and computational studies

Proton-bound dimers consisting of two glycerophospholipids with different headgroups were prepared using negative ion electrospray ionization and dissociated in a triple quadrupole mass spectrometer. Analysis of the tandem mass spectra of the dimers using the kinetic method provides, for the first time, an order of acidity for the phospholipid classes in the gas phase of PE < PA << PG < PS < PI. Hybrid density functional calculations on model phospholipids were used to predict the absolute deprotonation enthalpies of the phospholipid classes from isodesmic proton transfer reactions with phosphoric acid. The computational data largely support the experimental acidity trend, with the exception of the relative acidity ranking of the two most acidic phospholipid species. Possible causes of the discrepancy between experiment and theory are discussed and the experimental trend is recommended. The sequence of gas phase acidities for the phospholipid headgroups is found to (1) have little correlation with the relative ionization efficiencies of the phospholipid classes observed in the negative ion electrospray process, and (2) correlate well with fragmentation trends observed upon collisional activation of phospholipid \[M - H](-) anions. (c) 2005 American Society for Mass Spectrometry.

Bridging experiments, models and simulations : an integrative approach to validation in computational cardiac electrophysiology

Computational models in physiology often integrate functional and structural information from a large range of spatio-temporal scales from the ionic to the whole organ level. Their sophistication raises both expectations and scepticism concerning how computational methods can improve our understanding of living organisms and also how they can reduce, replace and refine animal experiments. A fundamental requirement to fulfil these expectations and achieve the full potential of computational physiology is a clear understanding of what models represent and how they can be validated. The present study aims at informing strategies for validation by elucidating the complex interrelations between experiments, models and simulations in cardiac electrophysiology. We describe the processes, data and knowledge involved in the construction of whole ventricular multiscale models of cardiac electrophysiology. Our analysis reveals that models, simulations, and experiments are intertwined, in an assemblage that is a system itself, namely the model-simulation-experiment (MSE) system. Validation must therefore take into account the complex interplay between models, simulations and experiments. Key points for developing strategies for validation are: 1) understanding sources of bio-variability is crucial to the comparison between simulation and experimental results; 2) robustness of techniques and tools is a pre-requisite to conducting physiological investigations using the MSE system; 3) definition and adoption of standards facilitates interoperability of experiments, models and simulations; 4) physiological validation must be understood as an iterative process that defines the specific aspects of electrophysiology the MSE system targets, and is driven by advancements in experimental and computational methods and the combination of both.

Dynamics of ribonuclease A and ribonuclease S: Computational and experimental studies

RNase S is a complex consisting of two proteolytic fragments of RNase A: the S peptide (residues 1-20) and S protein (residues 21-124). RNase S and RNase A have very similar X-ray structures and enzymatic activities. previous experiments have shown increased rates of hydrogen exchange and greater sensitivity to tryptic cleavage for RNase S relative to RNase A. It has therefore been asserted that the RNase S complex is considerably more dynamically flexible than RNase A. In the present study we examine the differences in the dynamics of RNaseS and RNase A computationally, by MD simulations, and experimentally, using trypsin cleavage as a probe of dynamics. The fluctuations around the average solution structure during the simulation were analyzed by measuring the RMS deviation in coordinates. No significant differences between RNase S and RNase A dynamics were observed in the simulations. We were able to account for the apparent discrepancy between simulation and experiment by a simple model, According to this model, the experimentally observed differences in dynamics can be quantitatively explained by the small amounts of free S peptide and S protein that are present in equilibrium with the RNase S complex. Thus, folded RNase A and the RNase S complex have identical dynamic behavior, despite the presence of a break in polypeptide chain between residues 20 and 21 in the latter molecule. This is in contrast to what has been widely believed for over 30 years about this important fragment complementation system.

Computational fluid dynamics simulations in aerosol and nucleation studies

Nucleation is the first step in the formation of a new phase inside a mother phase. Two main forms of nucleation can be distinguished. In homogeneous nucleation, the new phase is formed in a uniform substance. In heterogeneous nucleation, on the other hand, the new phase emerges on a pre-existing surface (nucleation site). Nucleation is the source of about 30% of all atmospheric aerosol which in turn has noticeable health effects and a significant impact on climate. Nucleation can be observed in the atmosphere, studied experimentally in the laboratory and is the subject of ongoing theoretical research. This thesis attempts to be a link between experiment and theory. By comparing simulation results to experimental data, the aim is to (i) better understand the experiments and (ii) determine where the theory needs improvement. Computational fluid dynamics (CFD) tools were used to simulate homogeneous onecomponent nucleation of n-alcohols in argon and helium as carrier gases, homogeneous nucleation in the water-sulfuric acid-system, and heterogeneous nucleation of water vapor on silver particles. In the nucleation of n-alcohols, vapor depletion, carrier gas effect and carrier gas pressure effect were evaluated, with a special focus on the pressure effect whose dependence on vapor and carrier gas properties could be specified. The investigation of nucleation in the water-sulfuric acid-system included a thorough analysis of the experimental setup, determining flow conditions, vapor losses, and nucleation zone. Experimental nucleation rates were compared to various theoretical approaches. We found that none of the considered theoretical descriptions of nucleation captured the role of water in the process at all relative humidities. Heterogeneous nucleation was studied in the activation of silver particles in a TSI 3785 particle counter which uses water as its working fluid. The role of the contact angle was investigated and the influence of incoming particle concentrations and homogeneous nucleation on counting efficiency determined.

Geant4 Hadronic Cascade Models and CMS Data Analysis : Computational Challenges in the LHC era

This work belongs to the field of computational high-energy physics (HEP). The key methods used in this thesis work to meet the challenges raised by the Large Hadron Collider (LHC) era experiments are object-orientation with software engineering, Monte Carlo simulation, the computer technology of clusters, and artificial neural networks. The first aspect discussed is the development of hadronic cascade models, used for the accurate simulation of medium-energy hadron-nucleus reactions, up to 10 GeV. These models are typically needed in hadronic calorimeter studies and in the estimation of radiation backgrounds. Various applications outside HEP include the medical field (such as hadron treatment simulations), space science (satellite shielding), and nuclear physics (spallation studies). Validation results are presented for several significant improvements released in Geant4 simulation tool, and the significance of the new models for computing in the Large Hadron Collider era is estimated. In particular, we estimate the ability of the Bertini cascade to simulate Compact Muon Solenoid (CMS) hadron calorimeter HCAL. LHC test beam activity has a tightly coupled cycle of simulation-to-data analysis. Typically, a Geant4 computer experiment is used to understand test beam measurements. Thus an another aspect of this thesis is a description of studies related to developing new CMS H2 test beam data analysis tools and performing data analysis on the basis of CMS Monte Carlo events. These events have been simulated in detail using Geant4 physics models, full CMS detector description, and event reconstruction. Using the ROOT data analysis framework we have developed an offline ANN-based approach to tag b-jets associated with heavy neutral Higgs particles, and we show that this kind of NN methodology can be successfully used to separate the Higgs signal from the background in the CMS experiment.

CAVVA: Computational Affective Video-in-Video Advertising

Advertising is ubiquitous in the online community and more so in the ever-growing and popular online video delivery websites (e. g., YouTube). Video advertising is becoming increasingly popular on these websites. In addition to the existing pre-roll/post-roll advertising and contextual advertising, this paper proposes an in-stream video advertising strategy-Computational Affective Video-in-Video Advertising (CAVVA). Humans being emotional creatures are driven by emotions as well as rational thought. We believe that emotions play a major role in influencing the buying behavior of users and hence propose a video advertising strategy which takes into account the emotional impact of the videos as well as advertisements. Given a video and a set of advertisements, we identify candidate advertisement insertion points (step 1) and also identify the suitable advertisements (step 2) according to theories from marketing and consumer psychology. We formulate this two part problem as a single optimization function in a non-linear 0-1 integer programming framework and provide a genetic algorithm based solution. We evaluate CAVVA using a subjective user-study and eye-tracking experiment. Through these experiments, we demonstrate that CAVVA achieves a good balance between the following seemingly conflicting goals of (a) minimizing the user disturbance because of advertisement insertion while (b) enhancing the user engagement with the advertising content. We compare our method with existing advertising strategies and show that CAVVA can enhance the user's experience and also help increase the monetization potential of the advertising content.

Neural and computational representations of decision variables

These studies explore how, where, and when representations of variables critical to decision-making are represented in the brain. In order to produce a decision, humans must first determine the relevant stimuli, actions, and possible outcomes before applying an algorithm that will select an action from those available. When choosing amongst alternative stimuli, the framework of value-based decision-making proposes that values are assigned to the stimuli and that these values are then compared in an abstract “value space” in order to produce a decision. Despite much progress, in particular regarding the pinpointing of ventromedial prefrontal cortex (vmPFC) as a region that encodes the value, many basic questions remain. In Chapter 2, I show that distributed BOLD signaling in vmPFC represents the value of stimuli under consideration in a manner that is independent of the type of stimulus it is. Thus the open question of whether value is represented in abstraction, a key tenet of value-based decision-making, is confirmed. However, I also show that stimulus-dependent value representations are also present in the brain during decision-making and suggest a potential neural pathway for stimulus-to-value transformations that integrates these two results.

More broadly speaking, there is both neural and behavioral evidence that two distinct control systems are at work during action selection. These two systems compose the “goal-directed system”, which selects actions based on an internal model of the environment, and the “habitual” system, which generates responses based on antecedent stimuli only. Computational characterizations of these two systems imply that they have different informational requirements in terms of input stimuli, actions, and possible outcomes. Associative learning theory predicts that the habitual system should utilize stimulus and action information only, while goal-directed behavior requires that outcomes as well as stimuli and actions be processed. In Chapter 3, I test whether areas of the brain hypothesized to be involved in habitual versus goal-directed control represent the corresponding theorized variables.

The question of whether one or both of these neural systems drives Pavlovian conditioning is less well-studied. Chapter 4 describes an experiment in which subjects were scanned while engaged in a Pavlovian task with a simple non-trivial structure. After comparing a variety of model-based and model-free learning algorithms (thought to underpin goal-directed and habitual decision-making, respectively), it was found that subjects’ reaction times were better explained by a model-based system. In addition, neural signaling of precision, a variable based on a representation of a world model, was found in the amygdala. These data indicate that the influence of model-based representations of the environment can extend even to the most basic learning processes.

Knowledge of the state of hidden variables in an environment is required for optimal inference regarding the abstract decision structure of a given environment and therefore can be crucial to decision-making in a wide range of situations. Inferring the state of an abstract variable requires the generation and manipulation of an internal representation of beliefs over the values of the hidden variable. In Chapter 5, I describe behavioral and neural results regarding the learning strategies employed by human subjects in a hierarchical state-estimation task. In particular, a comprehensive model fit and comparison process pointed to the use of "belief thresholding". This implies that subjects tended to eliminate low-probability hypotheses regarding the state of the environment from their internal model and ceased to update the corresponding variables. Thus, in concert with incremental Bayesian learning, humans explicitly manipulate their internal model of the generative process during hierarchical inference consistent with a serial hypothesis testing strategy.

Quantum Mechanics Studies of Fuel Cell Catalysts and Proton Conducting Ceramics with Validation by Experiment

We carried out quantum mechanics (QM) studies aimed at improving the performance of hydrogen fuel cells. This led to predictions of improved materials, some of which were subsequently validated with experiments by our collaborators.

In part I, the challenge was to find a replacement for the Pt cathode that would lead to improved performance for the Oxygen Reduction Reaction (ORR) while remaining stable under operational conditions and decreasing cost. Our design strategy was to find an alloy with composition Pt3M that would lead to surface segregation such that the top layer would be pure Pt, with the second and subsequent layers richer in M. Under operating conditions we expect the surface to have significant O and/or OH chemisorbed on the surface, and hence we searched for M that would remain segregated under these conditions. Using QM we examined surface segregation for 28 Pt₃M alloys, where M is a transition metal. We found that only Pt₃Os and Pt₃Ir showed significant surface segregation when O and OH are chemisorbed on the catalyst surfaces. This result indicates that Pt₃Os and Pt3Ir favor formation of a Pt-skin surface layer structure that would resist the acidic electrolyte corrosion during fuel cell operation environments. We chose to focus on Os because the phase diagram for Pt-Ir indicated that Pt-Ir could not form a homogeneous alloy at lower temperature. To determine the performance for ORR, we used QM to examine all intermediates, reaction pathways, and reaction barriers involved in the processes for which protons from the anode reactions react with O₂ to form H₂O. These QM calculations used our Poisson-Boltzmann implicit solvation model include the effects of the solvent (water with dielectric constant 78 with pH 7 at 298K). We found that the rate determination step (RDS) was the O_ad hydration reaction (O_ad + H₂O_ad -> OH_ad + OH_ad) in both cases, but that the barrier for pure Pt of 0.50 eV is reduced to 0.48 eV for Pt₃Os, which at 80 degrees C would increase the rate by 218%. We collaborated with the Pu-Wei Wu’s group to carry out experiments, where we found that the dealloying process-treated Pt2Os catalyst showed two-fold higher activity at 25 degrees C than pure Pt and that the alloy had 272% improved stability, validating our theoretical predictions.

We also carried out similar QM studies followed by experimental validation for the Os/Pt core-shell catalyst fabricated by the underpotential deposition (UPD) method. The QM results indicated that the RDS for ORR is a compromise between the OOH formation step (0.37 eV for Pt, 0.23 eV for Pt_2ML/Os core-shell) and H₂O formation steps (0.32 eV for Pt, 0.22 eV for Pt_2ML/Os core-shell). We found that Pt_2ML/Os has the highest activity (compared to pure Pt and to the Pt₃Os alloy) because the 0.37 eV barrier decreases to 0.23 eV. To understand what aspects of the core shell structure lead to this improved performance, we considered the effect on ORR of compressing the alloy slab to the dimensions of pure Pt. However this had little effect, with the same RDS barrier 0.37 eV. This shows that the ligand effect (the electronic structure modification resulting from the Os substrate) plays a more important role than the strain effect, and is responsible for the improved activity of the core- shell catalyst. Experimental materials characterization proves the core-shell feature of our catalyst. The electrochemical experiment for Pt_2ML/Os/C showed 3.5 to 5 times better ORR activity at 0.9V (vs. NHE) in 0.1M HClO₄ solution at 25 degrees C as compared to those of commercially available Pt/C. The excellent correlation between experimental half potential and the OH binding energies and RDS barriers validate the feasibility of predicting catalyst activity using QM calculation and a simple Langmuir–Hinshelwood model.

In part II, we used QM calculations to study methane stream reforming on a Ni-alloy catalyst surfaces for solid oxide fuel cell (SOFC) application. SOFC has wide fuel adaptability but the coking and sulfur poisoning will reduce its stability. Experimental results suggested that the Ni4Fe alloy improves both its activity and stability compared to pure Ni. To understand the atomistic origin of this, we carried out QM calculations on surface segregation and found that the most stable configuration for Ni₄Fe has a Fe atom distribution of (0%, 50%, 25%, 25%, 0%) starting at the bottom layer. We calculated that the binding of C atoms on the Ni4Fe surface is 142.9 Kcal/mol, which is about 10 Kcal/mol weaker compared to the pure Ni surface. This weaker C binding energy is expected to make coke formation less favorable, explaining why Ni₄Fe has better coking resistance. This result confirms the experimental observation. The reaction energy barriers for CHx decomposition and C binding on various alloy surface, Ni₄X (X=Fe, Co, Mn, and Mo), showed Ni₄Fe, Ni₄Co, and Fe₄Mn all have better coking resistance than pure Ni, but that only Ni₄Fe and Fe₄Mn have (slightly) improved activity compared to pure Ni.

In part III, we used QM to examine the proton transport in doped perovskite-ceramics. Here we used a 2x2x2 supercell of perovskite with composition Ba₈X₇M₁(OH)₁O₂₃ where X=Ce or Zr and M=Y, Gd, or Dy. Thus in each case a 4⁺ X is replace by a 3⁺ M plus a proton on one O. Here we predicted the barriers for proton diffusion allowing both includes intra-octahedron and inter-octahedra proton transfer. Without any restriction, we only observed the inter-octahedra proton transfer with similar energy barrier as previous computational work but 0.2 eV higher than experimental result for Y doped zirconate. For one restriction in our calculations is that the O_donor-O_acceptor atoms were kept at fixed distances, we found that the barrier difference between cerates/zirconates with various dopants are only 0.02~0.03 eV. To fully address performance one would need to examine proton transfer at grain boundaries, which will require larger scale ReaxFF reactive dynamics for systems with millions of atoms. The QM calculations used here will be used to train the ReaxFF force field.