"Making Sense of Figures": Statistics, Computing and Information Technologies in Agriculture and Biology in Britain, 1920s-1960s

Throughout the twentieth century statistical methods have increasingly become part of experimental research. In particular, statistics has made quantification processes meaningful in the soft sciences, which had traditionally relied on activities such as collecting and describing diversity rather than timing variation. The thesis explores this change in relation to agriculture and biology, focusing on analysis of variance and experimental design, the statistical methods developed by the mathematician and geneticist Ronald Aylmer Fisher during the 1920s. The role that Fisher’s methods acquired as tools of scientific research, side by side with the laboratory equipment and the field practices adopted by research workers, is here investigated bottom-up, beginning with the computing instruments and the information technologies that were the tools of the trade for statisticians. Four case studies show under several perspectives the interaction of statistics, computing and information technologies, giving on the one hand an overview of the main tools – mechanical calculators, statistical tables, punched and index cards, standardised forms, digital computers – adopted in the period, and on the other pointing out how these tools complemented each other and were instrumental for the development and dissemination of analysis of variance and experimental design. The period considered is the half-century from the early 1920s to the late 1960s, the institutions investigated are Rothamsted Experimental Station and the Galton Laboratory, and the statisticians examined are Ronald Fisher and Frank Yates.

Coordinazione embodied vs. Disembodied: TuCSon on cloud

Lo scopo della tesi è esplorare il nuovo dualismo tra calcolo situato e calcolo come mero servizio immateriale che si osserva nel rafforzarsi di due paradigmi apparentemente antitetici come Cloud Computing e Pervasive Computing. Si vuole quindi dimostrare che i due paradigmi sono complementari, e possibilmente sviluppare un modello e un approccio metodologico per sistemi distribuiti che sfrutti opportunamente le caratteristiche dei due paradigmi. A tale scopo si utilizzerà come caso di studio il modello TuCSoN con linguaggio ReSpecT, combinando opportunamente Situated ReSpecT con il modello Coordination as a Service (CaaS) espresso da TuCSoN on Cloud.

Model predictive control in thermal management of multiprocessor systems-on-chip

MultiProcessor Systems-on-Chip (MPSoC) are the core of nowadays and next generation computing platforms. Their relevance in the global market continuously increase, occupying an important role both in everydaylife products (e.g. smartphones, tablets, laptops, cars) and in strategical market sectors as aviation, defense, robotics, medicine. Despite of the incredible performance improvements in the recent years processors manufacturers have had to deal with issues, commonly called “Walls”, that have hindered the processors development. After the famous “Power Wall”, that limited the maximum frequency of a single core and marked the birth of the modern multiprocessors system-on-chip, the “Thermal Wall” and the “Utilization Wall” are the actual key limiter for performance improvements. The former concerns the damaging effects of the high temperature on the chip caused by the large power densities dissipation, whereas the second refers to the impossibility of fully exploiting the computing power of the processor due to the limitations on power and temperature budgets. In this thesis we faced these challenges by developing efficient and reliable solutions able to maximize performance while limiting the maximum temperature below a fixed critical threshold and saving energy. This has been possible by exploiting the Model Predictive Controller (MPC) paradigm that solves an optimization problem subject to constraints in order to find the optimal control decisions for the future interval. A fully-distributedMPC-based thermal controller with a far lower complexity respect to a centralized one has been developed. The control feasibility and interesting properties for the simplification of the control design has been proved by studying a partial differential equation thermal model. Finally, the controller has been efficiently included in more complex control schemes able to minimize energy consumption and deal with mixed-criticalities tasks

Implementation of network entropy algorithms on hpc machines, with application to high-dimensional experimental data

Network Theory is a prolific and lively field, especially when it approaches Biology. New concepts from this theory find application in areas where extensive datasets are already available for analysis, without the need to invest money to collect them. The only tools that are necessary to accomplish an analysis are easily accessible: a computing machine and a good algorithm. As these two tools progress, thanks to technology advancement and human efforts, wider and wider datasets can be analysed. The aim of this paper is twofold. Firstly, to provide an overview of one of these concepts, which originates at the meeting point between Network Theory and Statistical Mechanics: the entropy of a network ensemble. This quantity has been described from different angles in the literature. Our approach tries to be a synthesis of the different points of view. The second part of the work is devoted to presenting a parallel algorithm that can evaluate this quantity over an extensive dataset. Eventually, the algorithm will also be used to analyse high-throughput data coming from biology.

Maternal Smoking Impact on the Delivery Cost: a population-based study in the Emilia-Romagna region

This doctoral thesis is devoted to the study of the causal effects of the maternal smoking on the delivery cost. The interest of economic consequences of smoking in pregnancy have been studied fairly extensively in the USA, and very little is known in European context. To identify the causal relation between different maternal smoking status and the delivery cost in the Emilia-Romagna region two distinct methods were used. The first - geometric multidimensional - is mainly based on the multivariate approach and involves computing and testing the global imbalance, classifying cases in order to generate well-matched comparison groups, and then computing treatment effects. The second - structural modelling - refers to a general methodological account of model-building and model-testing. The main idea of this approach is to decompose the global mechanism into sub-mechanisms though a recursive decomposition of a multivariate distribution.

Parallel Architectures for Many-Core Systems-On-Chip in Deep Sub-Micron Technology

Despite the several issues faced in the past, the evolutionary trend of silicon has kept its constant pace. Today an ever increasing number of cores is integrated onto the same die. Unfortunately, the extraordinary performance achievable by the many-core paradigm is limited by several factors. Memory bandwidth limitation, combined with inefficient synchronization mechanisms, can severely overcome the potential computation capabilities. Moreover, the huge HW/SW design space requires accurate and flexible tools to perform architectural explorations and validation of design choices. In this thesis we focus on the aforementioned aspects: a flexible and accurate Virtual Platform has been developed, targeting a reference many-core architecture. Such tool has been used to perform architectural explorations, focusing on instruction caching architecture and hybrid HW/SW synchronization mechanism. Beside architectural implications, another issue of embedded systems is considered: energy efficiency. Near Threshold Computing is a key research area in the Ultra-Low-Power domain, as it promises a tenfold improvement in energy efficiency compared to super-threshold operation and it mitigates thermal bottlenecks. The physical implications of modern deep sub-micron technology are severely limiting performance and reliability of modern designs. Reliability becomes a major obstacle when operating in NTC, especially memory operation becomes unreliable and can compromise system correctness. In the present work a novel hybrid memory architecture is devised to overcome reliability issues and at the same time improve energy efficiency by means of aggressive voltage scaling when allowed by workload requirements. Variability is another great drawback of near-threshold operation. The greatly increased sensitivity to threshold voltage variations in today a major concern for electronic devices. We introduce a variation-tolerant extension of the baseline many-core architecture. By means of micro-architectural knobs and a lightweight runtime control unit, the baseline architecture becomes dynamically tolerant to variations.

Heterogeneous Multi-core Architectures for High Performance Computing

This thesis deals with heterogeneous architectures in standard workstations. Heterogeneous architectures represent an appealing alternative to traditional supercomputers because they are based on commodity components fabricated in large quantities. Hence their price-performance ratio is unparalleled in the world of high performance computing (HPC). In particular, different aspects related to the performance and consumption of heterogeneous architectures have been explored. The thesis initially focuses on an efficient implementation of a parallel application, where the execution time is dominated by an high number of floating point instructions. Then the thesis touches the central problem of efficient management of power peaks in heterogeneous computing systems. Finally it discusses a memory-bounded problem, where the execution time is dominated by the memory latency. Specifically, the following main contributions have been carried out: A novel framework for the design and analysis of solar field for Central Receiver Systems (CRS) has been developed. The implementation based on desktop workstation equipped with multiple Graphics Processing Units (GPUs) is motivated by the need to have an accurate and fast simulation environment for studying mirror imperfection and non-planar geometries. Secondly, a power-aware scheduling algorithm on heterogeneous CPU-GPU architectures, based on an efficient distribution of the computing workload to the resources, has been realized. The scheduler manages the resources of several computing nodes with a view to reducing the peak power. The two main contributions of this work follow: the approach reduces the supply cost due to high peak power whilst having negligible impact on the parallelism of computational nodes. from another point of view the developed model allows designer to increase the number of cores without increasing the capacity of the power supply unit. Finally, an implementation for efficient graph exploration on reconfigurable architectures is presented. The purpose is to accelerate graph exploration, reducing the number of random memory accesses.

The twofold role of Cloud Computing in Digital Forensics: target of investigations and helping hand to evidence analysis

This PhD thesis discusses the impact of Cloud Computing infrastructures on Digital Forensics in the twofold role of target of investigations and as a helping hand to investigators. The Cloud offers a cheap and almost limitless computing power and storage space for data which can be leveraged to commit either new or old crimes and host related traces. Conversely, the Cloud can help forensic examiners to find clues better and earlier than traditional analysis applications, thanks to its dramatically improved evidence processing capabilities. In both cases, a new arsenal of software tools needs to be made available. The development of this novel weaponry and its technical and legal implications from the point of view of repeatability of technical assessments is discussed throughout the following pages and constitutes the unprecedented contribution of this work

Implementazione dell'algoritmo filtered back-projection (FBP) per architetture low-power di tipo systems-on-chip

Il presente lavoro di tesi, svolto presso i laboratori dell'X-ray Imaging Group del Dipartimento di Fisica e Astronomia dell'Università di Bologna e all'interno del progetto della V Commissione Scientifica Nazionale dell'INFN, COSA (Computing on SoC Architectures), ha come obiettivo il porting e l’analisi di un codice di ricostruzione tomografica su architetture GPU installate su System-On-Chip low-power, al fine di sviluppare un metodo portatile, economico e relativamente veloce. Dall'analisi computazionale sono state sviluppate tre diverse versioni del porting in CUDA C: nella prima ci si è limitati a trasporre la parte più onerosa del calcolo sulla scheda grafica, nella seconda si sfrutta la velocità del calcolo matriciale propria del coprocessore (facendo coincidere ogni pixel con una singola unità di calcolo parallelo), mentre la terza è un miglioramento della precedente versione ottimizzata ulteriormente. La terza versione è quella definitiva scelta perché è la più performante sia dal punto di vista del tempo di ricostruzione della singola slice sia a livello di risparmio energetico. Il porting sviluppato è stato confrontato con altre due parallelizzazioni in OpenMP ed MPI. Si è studiato quindi, sia su cluster HPC, sia su cluster SoC low-power (utilizzando in particolare la scheda quad-core Tegra K1), l’efficienza di ogni paradigma in funzione della velocità di calcolo e dell’energia impiegata. La soluzione da noi proposta prevede la combinazione del porting in OpenMP e di quello in CUDA C. Tre core CPU vengono riservati per l'esecuzione del codice in OpenMP, il quarto per gestire la GPU usando il porting in CUDA C. Questa doppia parallelizzazione ha la massima efficienza in funzione della potenza e dell’energia, mentre il cluster HPC ha la massima efficienza in velocità di calcolo. Il metodo proposto quindi permetterebbe di sfruttare quasi completamente le potenzialità della CPU e GPU con un costo molto contenuto. Una possibile ottimizzazione futura potrebbe prevedere la ricostruzione di due slice contemporaneamente sulla GPU, raddoppiando circa la velocità totale e sfruttando al meglio l’hardware. Questo studio ha dato risultati molto soddisfacenti, infatti, è possibile con solo tre schede TK1 eguagliare e forse a superare, in seguito, la potenza di calcolo di un server tradizionale con il vantaggio aggiunto di avere un sistema portatile, a basso consumo e costo. Questa ricerca si va a porre nell’ambito del computing come uno tra i primi studi effettivi su architetture SoC low-power e sul loro impiego in ambito scientifico, con risultati molto promettenti.

Deep learning for computer vision: a comparison between convolutional neural networks and hierarchical temporal memories on object recognition tasks

In recent years, Deep Learning techniques have shown to perform well on a large variety of problems both in Computer Vision and Natural Language Processing, reaching and often surpassing the state of the art on many tasks. The rise of deep learning is also revolutionizing the entire field of Machine Learning and Pattern Recognition pushing forward the concepts of automatic feature extraction and unsupervised learning in general. However, despite the strong success both in science and business, deep learning has its own limitations. It is often questioned if such techniques are only some kind of brute-force statistical approaches and if they can only work in the context of High Performance Computing with tons of data. Another important question is whether they are really biologically inspired, as claimed in certain cases, and if they can scale well in terms of "intelligence". The dissertation is focused on trying to answer these key questions in the context of Computer Vision and, in particular, Object Recognition, a task that has been heavily revolutionized by recent advances in the field. Practically speaking, these answers are based on an exhaustive comparison between two, very different, deep learning techniques on the aforementioned task: Convolutional Neural Network (CNN) and Hierarchical Temporal memory (HTM). They stand for two different approaches and points of view within the big hat of deep learning and are the best choices to understand and point out strengths and weaknesses of each of them. CNN is considered one of the most classic and powerful supervised methods used today in machine learning and pattern recognition, especially in object recognition. CNNs are well received and accepted by the scientific community and are already deployed in large corporation like Google and Facebook for solving face recognition and image auto-tagging problems. HTM, on the other hand, is known as a new emerging paradigm and a new meanly-unsupervised method, that is more biologically inspired. It tries to gain more insights from the computational neuroscience community in order to incorporate concepts like time, context and attention during the learning process which are typical of the human brain. In the end, the thesis is supposed to prove that in certain cases, with a lower quantity of data, HTM can outperform CNN.

Scalable and Seamless Discovery and Selection of Services in Mobile Cloud Computing

Mobile devices are now capable of supporting a wide range of applications, many of which demand an ever increasing computational power. To this end, mobile cloud computing (MCC) has been proposed to address the limited computation power, memory, storage, and energy of such devices. An important challenge in MCC is to guarantee seamless discovery of services. To this end, this thesis proposes an architecture that provides user-transparent and low-latency service discovery, as well as automated service selection. Experimental results on a real cloud computing testbed demonstrated that the proposed work outperforms state of-the-art approaches by achieving extremely low discovery delay.

Distributed allocation and scheduling of applications on HPC platforms

High Performance Computing e una tecnologia usata dai cluster computazionali per creare sistemi di elaborazione che sono in grado di fornire servizi molto piu potenti rispetto ai computer tradizionali. Di conseguenza la tecnologia HPC e diventata un fattore determinante nella competizione industriale e nella ricerca. I sistemi HPC continuano a crescere in termini di nodi e core. Le previsioni indicano che il numero dei nodi arrivera a un milione a breve. Questo tipo di architettura presenta anche dei costi molto alti in termini del consumo delle risorse, che diventano insostenibili per il mercato industriale. Un scheduler centralizzato non e in grado di gestire un numero di risorse cosi alto, mantenendo un tempo di risposta ragionevole. In questa tesi viene presentato un modello di scheduling distribuito che si basa sulla programmazione a vincoli e che modella il problema dello scheduling grazie a una serie di vincoli temporali e vincoli sulle risorse che devono essere soddisfatti. Lo scheduler cerca di ottimizzare le performance delle risorse e tende ad avvicinarsi a un profilo di consumo desiderato, considerato ottimale. Vengono analizzati vari modelli diversi e ognuno di questi viene testato in vari ambienti.

A Trustability Metric for Code Search based on Developer Karma

The promise of search-driven development is that developers will save time and resources by reusing external code in their local projects. To efficiently integrate this code, users must be able to trust it, thus trustability of code search results is just as important as their relevance. In this paper, we introduce a trustability metric to help users assess the quality of code search results and therefore ease the cost-benefit analysis they undertake trying to find suitable integration candidates. The proposed trustability metric incorporates both user votes and cross-project activity of developers to calculate a "karma" value for each developer. Through the karma value of all its developers a project is ranked on a trustability scale. We present JBENDER, a proof-of-concept code search engine which implements our trustability metric and we discuss preliminary results from an evaluation of the prototype.

Towards Energy Consumption Measurement in a Cloud Computing Wireless Testbed

The evolution of the Next Generation Networks, especially the wireless broadband access technologies such as Long Term Evolution (LTE) and Worldwide Interoperability for Microwave Access (WiMAX), have increased the number of "all-IP" networks across the world. The enhanced capabilities of these access networks has spearheaded the cloud computing paradigm, where the end-users aim at having the services accessible anytime and anywhere. The services availability is also related with the end-user device, where one of the major constraints is the battery lifetime. Therefore, it is necessary to assess and minimize the energy consumed by the end-user devices, given its significance for the user perceived quality of the cloud computing services. In this paper, an empirical methodology to measure network interfaces energy consumption is proposed. By employing this methodology, an experimental evaluation of energy consumption in three different cloud computing access scenarios (including WiMAX) were performed. The empirical results obtained show the impact of accurate network interface states management and application network level design in the energy consumption. Additionally, the achieved outcomes can be used in further software-based models to optimized energy consumption, and increase the Quality of Experience (QoE) perceived by the end-users.