Balancing Energy and Performance in Dense Linear System Solvers for Hybrid ARM+GPU platforms

The high performance computing community has traditionally focused uniquely on the reduction of execution time, though in the last years, the optimization of energy consumption has become a main issue. A reduction of energy usage without a degradation of performance requires the adoption of energy-efficient hardware platforms accompanied by the development of energy-aware algorithms and computational kernels. The solution of linear systems is a key operation for many scientific and engineering problems. Its relevance has motivated an important amount of work, and consequently, it is possible to find high performance solvers for a wide variety of hardware platforms. In this work, we aim to develop a high performance and energy-efficient linear system solver. In particular, we develop two solvers for a low-power CPU-GPU platform, the NVIDIA Jetson TK1. These solvers implement the Gauss-Huard algorithm yielding an efficient usage of the target hardware as well as an efficient memory access. The experimental evaluation shows that the novel proposal reports important savings in both time and energy-consumption when compared with the state-of-the-art solvers of the platform.

An Adaptive and Hybrid Approach to Revisiting the Visibility Pipeline

We revisit the visibility problem, which is traditionally known in Computer Graphics and Vision fields as the process of computing a (potentially) visible set of primitives in the computational model of a scene. We propose a hybrid solution that uses a dry structure (in the sense of data reduction), a triangulation of the type , to accelerate the task of searching for visible primitives. We came up with a solution that is useful for real-time, on-line, interactive applications as 3D visualization. In such applications the main goal is to load the minimum amount of primitives from the scene during the rendering stage, as possible. For this purpose, our algorithm executes the culling by using a hybrid paradigm based on viewing-frustum, back-face culling and occlusion models. Results have shown substantial improvement over these traditional approaches if applied separately. This novel approach can be used in devices with no dedicated processors or with low processing power, as cell phones or embedded displays, or to visualize data through the Internet, as in virtual museums applications.

Hybrid Graph Heuristics within a Hyper-heuristic Approach to Exam Timetabling Problems

This paper is concerned with the hybridization of two graph coloring heuristics (Saturation Degree and Largest Degree), and their application within a hyperheuristic for exam timetabling problems. Hyper-heuristics can be seen as algorithms which intelligently select appropriate algorithms/heuristics for solving a problem. We developed a Tabu Search based hyper-heuristic to search for heuristic lists (of graph heuristics) for solving problems and investigated the heuristic lists found by employing knowledge discovery techniques. Two hybrid approaches (involving Saturation Degree and Largest Degree) including one which employs Case Based Reasoning are presented and discussed. Both the Tabu Search based hyper-heuristic and the hybrid approaches are tested on random and real-world exam timetabling problems. Experimental results are comparable with the best state-of-the-art approaches (as measured against established benchmark problems). The results also demonstrate an increased level of generality in our approach.

Soft computing based parameter identification in pavements and geomechanical systems

Accurate estimation of road pavement geometry and layer material properties through the use of proper nondestructive testing and sensor technologies is essential for evaluating pavement’s structural condition and determining options for maintenance and rehabilitation. For these purposes, pavement deflection basins produced by the nondestructive Falling Weight Deflectometer (FWD) test data are commonly used. The nondestructive FWD test drops weights on the pavement to simulate traffic loads and measures the created pavement deflection basins. Backcalculation of pavement geometry and layer properties using FWD deflections is a difficult inverse problem, and the solution with conventional mathematical methods is often challenging due to the ill-posed nature of the problem. In this dissertation, a hybrid algorithm was developed to seek robust and fast solutions to this inverse problem. The algorithm is based on soft computing techniques, mainly Artificial Neural Networks (ANNs) and Genetic Algorithms (GAs) as well as the use of numerical analysis techniques to properly simulate the geomechanical system. A widely used pavement layered analysis program ILLI-PAVE was employed in the analyses of flexible pavements of various pavement types; including full-depth asphalt and conventional flexible pavements, were built on either lime stabilized soils or untreated subgrade. Nonlinear properties of the subgrade soil and the base course aggregate as transportation geomaterials were also considered. A computer program, Soft Computing Based System Identifier or SOFTSYS, was developed. In SOFTSYS, ANNs were used as surrogate models to provide faster solutions of the nonlinear finite element program ILLI-PAVE. The deflections obtained from FWD tests in the field were matched with the predictions obtained from the numerical simulations to develop SOFTSYS models. The solution to the inverse problem for multi-layered pavements is computationally hard to achieve and is often not feasible due to field variability and quality of the collected data. The primary difficulty in the analysis arises from the substantial increase in the degree of non-uniqueness of the mapping from the pavement layer parameters to the FWD deflections. The insensitivity of some layer properties lowered SOFTSYS model performances. Still, SOFTSYS models were shown to work effectively with the synthetic data obtained from ILLI-PAVE finite element solutions. In general, SOFTSYS solutions very closely matched the ILLI-PAVE mechanistic pavement analysis results. For SOFTSYS validation, field collected FWD data were successfully used to predict pavement layer thicknesses and layer moduli of in-service flexible pavements. Some of the very promising SOFTSYS results indicated average absolute errors on the order of 2%, 7%, and 4% for the Hot Mix Asphalt (HMA) thickness estimation of full-depth asphalt pavements, full-depth pavements on lime stabilized soils and conventional flexible pavements, respectively. The field validations of SOFTSYS data also produced meaningful results. The thickness data obtained from Ground Penetrating Radar testing matched reasonably well with predictions from SOFTSYS models. The differences observed in the HMA and lime stabilized soil layer thicknesses observed were attributed to deflection data variability from FWD tests. The backcalculated asphalt concrete layer thickness results matched better in the case of full-depth asphalt flexible pavements built on lime stabilized soils compared to conventional flexible pavements. Overall, SOFTSYS was capable of producing reliable thickness estimates despite the variability of field constructed asphalt layer thicknesses.

Cyber foraging for green computing, improving performance and prolonging battery life of mobile devices

The increasing dependency of everyday life on mobile devices also increases the number and complexity of computing tasks to be supported by these devices. However, the inherent requirement of mobility restricts them from being resources rich both in terms of energy (battery capacity) and other computing resources such as processing capacity, memory and other resources. This thesis looks into cyber foraging technique of offloading computing tasks. Various experiments on android mobile devices are carried out to evaluate offloading benefits in terms of sustainability advantage, prolonging battery life and augmenting the performance of mobile devices. This thesis considers two scenarios of cyber foraging namely opportunistic offloading and competitive offloading. These results show that the offloading scenarios are important for both green computing and resource augmentation of mobile devices. A significant advantage in battery life gain and performance enhancement is obtained. Moreover, cyber foraging is proved to be efficient in minimizing energy consumption per computing tasks. The work is based on scavenger cyber foraging system. In addition, the work can be used as a basis for studying cyber foraging and other similar approaches such as mobile cloud/edge computing for internet of things devices and improving the user experiences of applications by minimizing latencies through the use of potential nearby surrogates.

Improving drug discovery using hybrid softcomputing methods

Virtual screening (VS) methods can considerably aid clinical research, predicting how ligands interact with drug targets. Most VS methods suppose a unique binding site for the target, but it has been demonstrated that diverse ligands interact with unrelated parts of the target and many VS methods do not take into account this relevant fact. This problem is circumvented by a novel VS methodology named BINDSURF that scans the whole protein surface in order to find new hotspots, where ligands might potentially interact with, and which is implemented in last generation massively parallel GPU hardware, allowing fast processing of large ligand databases. BINDSURF can thus be used in drug discovery, drug design, drug repurposing and therefore helps considerably in clinical research. However, the accuracy of most VS methods and concretely BINDSURF is constrained by limitations in the scoring function that describes biomolecular interactions, and even nowadays these uncertainties are not completely understood. In order to improve accuracy of the scoring functions used in BINDSURF we propose a hybrid novel approach where neural networks (NNET) and support vector machines (SVM) methods are trained with databases of known active (drugs) and inactive compounds, being this information exploited afterwards to improve BINDSURF VS predictions.

Multi-sensor Cloud and Aerosol Retrieval Simulator and Its Applications

Executing a cloud or aerosol physical properties retrieval algorithm from controlled synthetic data is an important step in retrieval algorithm development. Synthetic data can help answer questions about the sensitivity and performance of the algorithm or aid in determining how an existing retrieval algorithm may perform with a planned sensor. Synthetic data can also help in solving issues that may have surfaced in the retrieval results. Synthetic data become very important when other validation methods, such as field campaigns,are of limited scope. These tend to be of relatively short duration and often are costly. Ground stations have limited spatial coverage whilesynthetic data can cover large spatial and temporal scales and a wide variety of conditions at a low cost. In this work I develop an advanced cloud and aerosol retrieval simulator for the MODIS instrument, also known as Multi-sensor Cloud and Aerosol Retrieval Simulator (MCARS). In a close collaboration with the modeling community I have seamlessly combined the GEOS-5 global climate model with the DISORT radiative transfer code, widely used by the remote sensing community, with the observations from the MODIS instrument to create the simulator. With the MCARS simulator it was then possible to solve the long standing issue with the MODIS aerosol optical depth retrievals that had a low bias for smoke aerosols. MODIS aerosol retrieval did not account for effects of humidity on smoke aerosols. The MCARS simulator also revealed an issue that has not been recognized previously, namely,the value of fine mode fraction could create a linear dependence between retrieved aerosol optical depth and land surface reflectance. MCARS provided the ability to examine aerosol retrievals against “ground truth” for hundreds of thousands of simultaneous samples for an area covered by only three AERONET ground stations. Findings from MCARS are already being used to improve the performance of operational MODIS aerosol properties retrieval algorithms. The modeling community will use the MCARS data to create new parameterizations for aerosol properties as a function of properties of the atmospheric column and gain the ability to correct any assimilated retrieval data that may display similar dependencies in comparisons with ground measurements.

Collaborative filtering and deep learning based hybrid recommendation for cold start problem

Recommender systems (RS) are used by many social networking applications and online e-commercial services. Collaborative filtering (CF) is one of the most popular approaches used for RS. However traditional CF approach suffers from sparsity and cold start problems. In this paper, we propose a hybrid recommendation model to address the cold start problem, which explores the item content features learned from a deep learning neural network and applies them to the timeSVD++ CF model. Extensive experiments are run on a large Netflix rating dataset for movies. Experiment results show that the proposed hybrid recommendation model provides a good prediction for cold start items, and performs better than four existing recommendation models for rating of non-cold start items.

Impact of single parameter changes on Ceph cloud storage performance

In a general purpose cloud system efficiencies are yet to be had from supporting diverse applications and their requirements within a storage system used for a private cloud. Supporting such diverse requirements poses a significant challenge in a storage system that supports fine grained configuration on a variety of parameters. This paper uses the Ceph distributed file system, and in particular its global parameters, to show how a single changed parameter can effect the performance for a range of access patterns when tested with an OpenStack cloud system.

Flexible Computing Systems For AI Acceleration At The Extreme Edge Of The IoT

Embedding intelligence in extreme edge devices allows distilling raw data acquired from sensors into actionable information, directly on IoT end-nodes. This computing paradigm, in which end-nodes no longer depend entirely on the Cloud, offers undeniable benefits, driving a large research area (TinyML) to deploy leading Machine Learning (ML) algorithms on micro-controller class of devices. To fit the limited memory storage capability of these tiny platforms, full-precision Deep Neural Networks (DNNs) are compressed by representing their data down to byte and sub-byte formats, in the integer domain. However, the current generation of micro-controller systems can barely cope with the computing requirements of QNNs. This thesis tackles the challenge from many perspectives, presenting solutions both at software and hardware levels, exploiting parallelism, heterogeneity and software programmability to guarantee high flexibility and high energy-performance proportionality. The first contribution, PULP-NN, is an optimized software computing library for QNN inference on parallel ultra-low-power (PULP) clusters of RISC-V processors, showing one order of magnitude improvements in performance and energy efficiency, compared to current State-of-the-Art (SoA) STM32 micro-controller systems (MCUs) based on ARM Cortex-M cores. The second contribution is XpulpNN, a set of RISC-V domain specific instruction set architecture (ISA) extensions to deal with sub-byte integer arithmetic computation. The solution, including the ISA extensions and the micro-architecture to support them, achieves energy efficiency comparable with dedicated DNN accelerators and surpasses the efficiency of SoA ARM Cortex-M based MCUs, such as the low-end STM32M4 and the high-end STM32H7 devices, by up to three orders of magnitude. To overcome the Von Neumann bottleneck while guaranteeing the highest flexibility, the final contribution integrates an Analog In-Memory Computing accelerator into the PULP cluster, creating a fully programmable heterogeneous fabric that demonstrates end-to-end inference capabilities of SoA MobileNetV2 models, showing two orders of magnitude performance improvements over current SoA analog/digital solutions.

Adaptable security in the cloud-to-thing continuum

Recent technological advancements have played a key role in seamlessly integrating cloud, edge, and Internet of Things (IoT) technologies, giving rise to the Cloud-to-Thing Continuum paradigm. This cloud model connects many heterogeneous resources that generate a large amount of data and collaborate to deliver next-generation services. While it has the potential to reshape several application domains, the number of connected entities remarkably broadens the security attack surface. One of the main problems is the lack of security measures to adapt to the dynamic and evolving conditions of the Cloud-To-Thing Continuum. To address this challenge, this dissertation proposes novel adaptable security mechanisms. Adaptable security is the capability of security controls, systems, and protocols to dynamically adjust to changing conditions and scenarios. However, since the design and development of novel security mechanisms can be explored from different perspectives and levels, we place our attention on threat modeling and access control. The contributions of the thesis can be summarized as follows. First, we introduce a model-based methodology that secures the design of edge and cyber-physical systems. This solution identifies threats, security controls, and moving target defense techniques based on system features. Then, we focus on access control management. Since access control policies are subject to modifications, we evaluate how they can be efficiently shared among distributed areas, highlighting the effectiveness of distributed ledger technologies. Furthermore, we propose a risk-based authorization middleware, adjusting permissions based on real-time data, and a federated learning framework that enhances trustworthiness by weighting each client's contributions according to the quality of their partial models. Finally, since authorization revocation is another critical concern, we present an efficient revocation scheme for verifiable credentials in IoT networks, featuring decentralization, demanding minimum storage and computing capabilities. All the mechanisms have been evaluated in different conditions, proving their adaptability to the Cloud-to-Thing Continuum landscape.

Integrating mobile devices into a scalable Grid Computing architecture designed to perform distributed computations

The idea of Grid Computing originated in the nineties and found its concrete applications in contexts like the SETI@home project where a lot of computers (offered by volunteers) cooperated, performing distributed computations, inside the Grid environment analyzing radio signals trying to find extraterrestrial life. The Grid was composed of traditional personal computers but, with the emergence of the first mobile devices like Personal Digital Assistants (PDAs), researchers started theorizing the inclusion of mobile devices into Grid Computing; although impressive theoretical work was done, the idea was discarded due to the limitations (mainly technological) of mobile devices available at the time. Decades have passed, and now mobile devices are extremely more performant and numerous than before, leaving a great amount of resources available on mobile devices, such as smartphones and tablets, untapped. Here we propose a solution for performing distributed computations over a Grid Computing environment that utilizes both desktop and mobile devices, exploiting the resources from day-to-day mobile users that alternatively would end up unused. The work starts with an introduction on what Grid Computing is, the evolution of mobile devices, the idea of integrating such devices into the Grid and how to convince device owners to participate in the Grid. Then, the tone becomes more technical, starting with an explanation on how Grid Computing actually works, followed by the technical challenges of integrating mobile devices into the Grid. Next, the model, which constitutes the solution offered by this study, is explained, followed by a chapter regarding the realization of a prototype that proves the feasibility of distributed computations over a Grid composed by both mobile and desktop devices. To conclude future developments and ideas to improve this project are presented.

Design, Synthesis And Biological Evaluation Of Hybrid Bioisoster Derivatives Of N-acylhydrazone And Furoxan Groups With Potential And Selective Anti-trypanosoma Cruzi Activity.

Hybrid bioisoster derivatives from N-acylhydrazones and furoxan groups were designed with the objective of obtaining at least a dual mechanism of action: cruzain inhibition and nitric oxide (NO) releasing activity. Fifteen designed compounds were synthesized varying the substitution in N-acylhydrazone and in furoxan group as well. They had its anti-Trypanosoma cruzi activity in amastigotes forms, NO releasing potential and inhibitory cruzain activity evaluated. The two most active compounds (6, 14) both in the parasite amastigotes and in the enzyme contain the nitro group in para position of the aromatic ring. The permeability screening in Caco-2 cell and cytotoxicity assay in human cells were performed for those most active compounds and both showed to be less cytotoxic than the reference drug, benznidazole. Compound 6 was the most promising, since besides activity it showed good permeability and selectivity index, higher than the reference drug. Thereby the compound 6 was considered as a possible candidate for additional studies.

Low-density Three-dimensional Foam Using Self-reinforced Hybrid Two-dimensional Atomic Layers.

Low-density nanostructured foams are often limited in applications due to their low mechanical and thermal stabilities. Here we report an approach of building the structural units of three-dimensional (3D) foams using hybrid two-dimensional (2D) atomic layers made of stacked graphene oxide layers reinforced with conformal hexagonal boron nitride (h-BN) platelets. The ultra-low density (1/400 times density of graphite) 3D porous structures are scalably synthesized using solution processing method. A layered 3D foam structure forms due to presence of h-BN and significant improvements in the mechanical properties are observed for the hybrid foam structures, over a range of temperatures, compared with pristine graphene oxide or reduced graphene oxide foams. It is found that domains of h-BN layers on the graphene oxide framework help to reinforce the 2D structural units, providing the observed improvement in mechanical integrity of the 3D foam structure.

Improved Selective Cholesterol Adsorption By Molecularly Imprinted Poly(methacrylic Acid)/silica (pmaa-sio₂) Hybrid Material Synthesized With Different Molar Ratios.

The present paper describes the synthesis of molecularly imprinted polymer - poly(methacrylic acid)/silica and reports its performance feasibility with desired adsorption capacity and selectivity for cholesterol extraction. Two imprinted hybrid materials were synthesized at different methacrylic acid (MAA)/tetraethoxysilane (TEOS) molar ratios (6:1 and 1:5) and characterized by FT-IR, TGA, SEM and textural data. Cholesterol adsorption on hybrid materials took place preferably in apolar solvent medium, especially in chloroform. From the kinetic data, the equilibrium time was reached quickly, being 12 and 20 min for the polymers synthesized at MAA/TEOS molar ratio of 6:1 and 1:5, respectively. The pseudo-second-order model provided the best fit for cholesterol adsorption on polymers, confirming the chemical nature of the adsorption process, while the dual-site Langmuir-Freundlich equation presented the best fit to the experimental data, suggesting the existence of two kinds of adsorption sites on both polymers. The maximum adsorption capacities obtained for the polymers synthesized at MAA/TEOS molar ratios of 6:1 and 1:5 were found to be 214.8 and 166.4 mg g(-1), respectively. The results from isotherm data also indicated higher adsorption capacity for both imprinted polymers regarding to corresponding non-imprinted polymers. Nevertheless, taking into account the retention parameters and selectivity of cholesterol in the presence of structurally analogue compounds (5-α-cholestane and 7-dehydrocholesterol), it was observed that the polymer synthesized at the MAA/TEOS molar ratio of 6:1 was much more selective for cholesterol than the one prepared at the ratio of 1:5, thus suggesting that selective binding sites ascribed to the carboxyl group from MAA play a central role in the imprinting effect created on MIP.