Energy consumption of parallel algorithms for solving linear systems on HPC architecture

Modern High-Performance Computing HPC systems are gradually increasing in size and complexity due to the correspondent demand of larger simulations requiring more complicated tasks and higher accuracy. However, as side effects of the Dennard’s scaling approaching its ultimate power limit, the efficiency of software plays also an important role in increasing the overall performance of a computation. Tools to measure application performance in these increasingly complex environments provide insights into the intricate ways in which software and hardware interact. The monitoring of the power consumption in order to save energy is possible through processors interfaces like Intel Running Average Power Limit RAPL. Given the low level of these interfaces, they are often paired with an application-level tool like Performance Application Programming Interface PAPI. Since several problems in many heterogeneous fields can be represented as a complex linear system, an optimized and scalable linear system solver algorithm can decrease significantly the time spent to compute its resolution. One of the most widely used algorithms deployed for the resolution of large simulation is the Gaussian Elimination, which has its most popular implementation for HPC systems in the Scalable Linear Algebra PACKage ScaLAPACK library. However, another relevant algorithm, which is increasing in popularity in the academic field, is the Inhibition Method. This thesis compares the energy consumption of the Inhibition Method and Gaussian Elimination from ScaLAPACK to profile their execution during the resolution of linear systems above the HPC architecture offered by CINECA. Moreover, it also collates the energy and power values for different ranks, nodes, and sockets configurations. The monitoring tools employed to track the energy consumption of these algorithms are PAPI and RAPL, that will be integrated with the parallel execution of the algorithms managed with the Message Passing Interface MPI.

Synthetic generation of amyloid PET images by non-linear dimensionality reduction inversion

Privacy issues and data scarcity in PET field call for efficient methods to expand datasets via synthetic generation of new data that cannot be traced back to real patients and that are also realistic. In this thesis, machine learning techniques were applied to 1001 amyloid-beta PET images, which had undergone a diagnosis of Alzheimer’s disease: the evaluations were 540 positive, 457 negative and 4 unknown. Isomap algorithm was used as a manifold learning method to reduce the dimensions of the PET dataset; a numerical scale-free interpolation method was applied to invert the dimensionality reduction map. The interpolant was tested on the PET images via LOOCV, where the removed images were compared with the reconstructed ones with the mean SSIM index (MSSIM = 0.76 ± 0.06). The effectiveness of this measure is questioned, since it indicated slightly higher performance for a method of comparison using PCA (MSSIM = 0.79 ± 0.06), which gave clearly poor quality reconstructed images with respect to those recovered by the numerical inverse mapping. Ten synthetic PET images were generated and, after having been mixed with ten originals, were sent to a team of clinicians for the visual assessment of their realism; no significant agreements were found either between clinicians and the true image labels or among the clinicians, meaning that original and synthetic images were indistinguishable. The future perspective of this thesis points to the improvement of the amyloid-beta PET research field by increasing available data, overcoming the constraints of data acquisition and privacy issues. Potential improvements can be achieved via refinements of the manifold learning and the inverse mapping stages during the PET image analysis, by exploring different combinations in the choice of algorithm parameters and by applying other non-linear dimensionality reduction algorithms. A final prospect of this work is the search for new methods to assess image reconstruction quality.

Robôs quadrúpedes: geração de trajectórias em tempo real usando sistemas dinâmicos não-lineares

A geração de trajectórias de robôs em tempo real é uma tarefa muito complexa, não existindo ainda um algoritmo que a permita resolver de forma eficaz. De facto, há controladores eficientes para trajectórias previamente definidas, todavia, a adaptação a variações imprevisíveis, como sendo terrenos irregulares ou obstáculos, constitui ainda um problema em aberto na geração de trajectórias em tempo real de robôs. Neste trabalho apresentam-se modelos de geradores centrais de padrões de locomoção (CPGs), inspirados na biologia, que geram os ritmos locomotores num robô quadrúpede. Os CPGs são modelados matematicamente por sistemas acoplados de células (ou neurónios), sendo a dinâmica de cada célula dada por um sistema de equações diferenciais ordinárias não lineares. Assume-se que as trajectórias dos robôs são constituídas por esta parte rítmica e por uma parte discreta. A parte discreta pode ser embebida na parte rítmica, (a.1) como um offset ou (a.2) adicionada às expressões rítmicas, ou (b) pode ser calculada independentemente e adicionada exactamente antes do envio dos sinais para as articulações do robô. A parte discreta permite inserir no passo locomotor uma perturbação, que poderá estar associada à locomoção em terrenos irregulares ou à existência de obstáculos na trajectória do robô. Para se proceder á análise do sistema com parte discreta, será variado o parâmetro g. O parâmetro g, presente nas equações da parte discreta, representa o offset do sinal após a inclusão da parte discreta. Revê-se a teoria de bifurcação e simetria que permite a classificação das soluções periódicas produzidas pelos modelos de CPGs com passos locomotores quadrúpedes. Nas simulações numéricas, usam-se as equações de Morris-Lecar e o oscilador de Hopf como modelos da dinâmica interna de cada célula para a parte rítmica. A parte discreta é modelada por um sistema inspirado no modelo VITE. Medem-se a amplitude e a frequência de dois passos locomotores para variação do parâmetro g, no intervalo [-5;5]. Consideram-se duas formas distintas de incluir a parte discreta na parte rítmica: (a) como um (a.1) offset ou (a.2) somada nas expressões que modelam a parte rítmica, e (b) somada ao sinal da parte rítmica antes de ser enviado às articulações do robô. No caso (a.1), considerando o oscilador de Hopf como dinâmica interna das células, verifica-se que a amplitude e frequência se mantêm constantes para -50.2. A extensão do movimento varia de forma directamente proporcional à amplitude. No caso das equações de Morris-Lecar, quando a componente discreta é embebida (a.2), a amplitude e a frequência aumentam e depois diminuem para - 0.170.5 Pode concluir-se que: (1) a melhor forma de inserção da parte discreta que menos perturbação insere no robô é a inserção como offset; (2) a inserção da parte discreta parece ser independente do sistema de equações diferenciais ordinárias que modelam a dinâmica interna de cada célula. Como trabalho futuro, é importante prosseguir o estudo das diferentes formas de inserção da parte discreta na parte rítmica do movimento, para que se possa gerar uma locomoção quadrúpede, robusta, flexível, com objectivos, em terrenos irregulares, modelada por correcções discretas aos padrões rítmicos.

Optimal tuning of fractional controllers using genetic algorithms

This study addresses the optimization of fractional algorithms for the discrete-time control of linear and non-linear systems. The paper starts by analyzing the fundamentals of fractional control systems and genetic algorithms. In a second phase the paper evaluates the problem in an optimization perspective. The results demonstrate the feasibility of the evolutionary strategy and the adaptability to distinct types of systems.

System Modeling and Control Through Fractional-Order Algorithms

The theory of fractional calculus goes back to the beginning of thr throry of differential calculus but its inherent complexity postponed the applications of the associated concepts. In the last decade the progress in the areas of chaos and fractals revealed subtle relationships with the fractional calculus leading to an increasing interest in the development of the new paradigm. In the area of automaticcontrol preliminary work has already been carried out but the proposed algorithms are restricted to the frequency domain. The paper discusses the design of fractional-order discrete-time controllers. The algorithms studied adopt the time domein, which makes them suited for z-transform analusis and discrete-time implementation. The performance of discrete-time fractional-order controllers with linear and non-linear systems is also investigated.

Local-global splitting for spatiotemporal-adaptive multiscale methods

We present a novel spatiotemporal-adaptive Multiscale Finite Volume (MsFV) method, which is based on the natural idea that the global coarse-scale problem has longer characteristic time than the local fine-scale problems. As a consequence, the global problem can be solved with larger time steps than the local problems. In contrast to the pressure-transport splitting usually employed in the standard MsFV approach, we propose to start directly with a local-global splitting that allows to locally retain the original degree of coupling. This is crucial for highly non-linear systems or in the presence of physical instabilities. To obtain an accurate and efficient algorithm, we devise new adaptive criteria for global update that are based on changes of coarse-scale quantities rather than on fine-scale quantities, as it is routinely done before in the adaptive MsFV method. By means of a complexity analysis we show that the adaptive approach gives a noticeable speed-up with respect to the standard MsFV algorithm. In particular, it is efficient in case of large upscaling factors, which is important for multiphysics problems. Based on the observation that local time stepping acts as a smoother, we devise a self-correcting algorithm which incorporates the information from previous times to improve the quality of the multiscale approximation. We present results of multiphase flow simulations both for Darcy-scale and multiphysics (hybrid) problems, in which a local pore-scale description is combined with a global Darcy-like description. The novel spatiotemporal-adaptive multiscale method based on the local-global splitting is not limited to porous media flow problems, but it can be extended to any system described by a set of conservation equations.

Substitution-Newton-Raphson method applied to the modeling of a vapour compression refrigeration system using different representations of the thermodynamic properties of R-134a

This paper gives a detailed presentation of the Substitution-Newton-Raphson method, suitable for large sparse non-linear systems. It combines the Successive Substitution method and the Newton-Raphson method in such way as to take the best advantages of both, keeping the convergence features of the Newton-Raphson with the low requirements of memory and time of the Successive Substitution schemes. The large system is solved employing few effective variables, using the greatest possible part of the model equations in substitution fashion to fix the remaining variables, but maintaining the convergence characteristics of the Newton-Raphson. The methodology is exemplified through a simple algebraic system, and applied to a simple thermodynamic, mechanical and heat transfer modeling of a single-stage vapor compression refrigeration system. Three distinct approaches for reproducing the thermodynamic properties of the refrigerant R-134a are compared: the linear interpolation from tabulated data, the use of polynomial fitted curves and the use of functions derived from the Helmholtz free energy.

Nonlinear system identification using particle swarm optimisation tuned radial basis function models

A novel particle swarm optimisation (PSO) tuned radial basis function (RBF) network model is proposed for identification of non-linear systems. At each stage of orthogonal forward regression (OFR) model construction process, PSO is adopted to tune one RBF unit's centre vector and diagonal covariance matrix by minimising the leave-one-out (LOO) mean square error (MSE). This PSO aided OFR automatically determines how many tunable RBF nodes are sufficient for modelling. Compared with the-state-of-the-art local regularisation assisted orthogonal least squares algorithm based on the LOO MSE criterion for constructing fixed-node RBF network models, the PSO tuned RBF model construction produces more parsimonious RBF models with better generalisation performance and is often more efficient in model construction. The effectiveness of the proposed PSO aided OFR algorithm for constructing tunable node RBF models is demonstrated using three real data sets.

Estudo e implementação de um controlador preditivo generalizado bilinear compensado adaptativo

The present work presents the study and implementation of an adaptive bilinear compensated generalized predictive controller. This work uses conventional techniques of predictive control and includes techniques of adaptive control for better results. In order to solve control problems frequently found in the chemical industry, bilinear models are considered to represent the dynamics of the studied systems. Bilinear models are simpler than general nonlinear model, however it can to represent the intrinsic not-linearities of industrial processes. The linearization of the model, by the approach to time step quasilinear , is used to allow the application of the equations of the generalized predictive controller (GPC). Such linearization, however, generates an error of prediction, which is minimized through a compensation term. The term in study is implemented in an adaptive form, due to the nonlinear relationship between the input signal and the prediction error.Simulation results show the efficiency of adaptive predictive bilinear controller in comparison with the conventional.

NUMERICAL SIMULATION of THE BEHAVIOR of THE BEHAVIOR of REINFORCED CONCRETE BEAMS and COLUMNS BY FEM CAASTEM 2000 PROGRAM

The objective of this paper is the numerical study of the behavior of reinforced concrete beams and columns by non-linear numerical simulations. The numerical analysis is based on the finite element method implemented in CASTEM 2000. This program uses the constitutive elastoplastic perfect model for the steel, the Drucker-Prager model for the concrete and the Newton-Raphson for the solution of non-linear systems. This work concentrates on the determination of equilibrium curves to the beams and force-strain curves to the columns. The numeric responses are confronted with experimental results found in the literature in order to check there liability of the numerical analyses.

On relaxed LMI-based design for fuzzy controllers

Relaxed conditions for stability of nonlinear continuous-time systems given by fuzzy models axe presented. A theoretical analysis shows that the proposed method provides better or at least the same results of the methods presented in the literature. Digital simulations exemplify this fact. This result is also used for fuzzy regulators design. The nonlinear systems are represented by fuzzy models proposed by Takagi and Sugeno. The stability analysis and the design of controllers axe described by LMIs (Linear Matrix Inequalities), that can be solved efficiently using convex programming techniques.

Espectroscopia no infravermelho próximo e métodos de calibração multivariada aplicados à determinação simultânea de parâmetros bioquímicos em plasma sanguíneo

In this work, the quantitative analysis of glucose, triglycerides and cholesterol (total and HDL) in both rat and human blood plasma was performed without any kind of pretreatment of samples, by using near infrared spectroscopy (NIR) combined with multivariate methods. For this purpose, different techniques and algorithms used to pre-process data, to select variables and to build multivariate regression models were compared between each other, such as partial least squares regression (PLS), non linear regression by artificial neural networks, interval partial least squares regression (iPLS), genetic algorithm (GA), successive projections algorithm (SPA), amongst others. Related to the determinations of rat blood plasma samples, the variables selection algorithms showed satisfactory results both for the correlation coefficients (R²) and for the values of root mean square error of prediction (RMSEP) for the three analytes, especially for triglycerides and cholesterol-HDL. The RMSEP values for glucose, triglycerides and cholesterol-HDL obtained through the best PLS model were 6.08, 16.07 e 2.03 mg dL-1, respectively. In the other case, for the determinations in human blood plasma, the predictions obtained by the PLS models provided unsatisfactory results with non linear tendency and presence of bias. Then, the ANN regression was applied as an alternative to PLS, considering its ability of modeling data from non linear systems. The root mean square error of monitoring (RMSEM) for glucose, triglycerides and total cholesterol, for the best ANN models, were 13.20, 10.31 e 12.35 mg dL-1, respectively. Statistical tests (F and t) suggest that NIR spectroscopy combined with multivariate regression methods (PLS and ANN) are capable to quantify the analytes (glucose, triglycerides and cholesterol) even when they are present in highly complex biological fluids, such as blood plasma

Cerâmicas eletrônicas à base de SnO2 e TiO2

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Power analysis applying the instantaneous complex power analytical expressions on a RL symmetrical three-phase system

This paper enhances some concepts of the Instantaneous Complex Power Theory by analyzing the analytical expressions for voltages, currents and powers developed on a symmetrical RL three-phase system, during the transient caused by a sinusoidal voltage excitation. The powers delivered to an ideal inductor will be interpreted, allowing a deep insight in the power phenomenon by analyzing the voltages in each element of the circuit. The results can be applied to the understanding of non-linear systems subject to sinusoidal voltage excitation and distorted currents.

A comparative study between two relaxed LMI-based fuzzy control designs

A comparative study, with theoretical analysis and digital simulations, of two conditions based on LMI for the quadratic stability of nonlinear continuous-time dynamic systems, described by Takagi-Sugeno fuzzy models, are presented. This paper shows that the methods proposed by Teixeira et. al. in 2003 provide better or at least the same results of a recent method presented in the literature. © 2005 IEEE.