No description available.

Submitted by Izabel Franco (izabel-franco@ufscar.br) on 2016-10-03T18:47:50Z No. of bitstreams: 1 DissMAA.pdf: 3325891 bytes, checksum: 1b4744d48d74943990bed42753cc4b4c (MD5) / Approved for entry into archive by Marina Freitas (marinapf@ufscar.br) on 2016-10-20T18:27:58Z (GMT) No. of bitstreams: 1 DissMAA.pdf: 3325891 bytes, checksum: 1b4744d48d74943990bed42753cc4b4c (MD5) / Approved for entry into archive by Marina Freitas (marinapf@ufscar.br) on 2016-10-20T18:28:04Z (GMT) No. of bitstreams: 1 DissMAA.pdf: 3325891 bytes, checksum: 1b4744d48d74943990bed42753cc4b4c (MD5) / Made available in DSpace on 2016-10-20T18:28:13Z (GMT). No. of bitstreams: 1 DissMAA.pdf: 3325891 bytes, checksum: 1b4744d48d74943990bed42753cc4b4c (MD5) Previous issue date: 2016-03-04 / Não recebi financiamento / The use of technology based on Field Programmable Gate Arrays (FPGAs), a reconfigurable technology, has become a frequent object of study. This technique is feasible and a promising application in the development of embedded systems, however, the difficulty in finding a flexible and efficient way to perform such an application is their bigger problem. In this work, a virtual and reconfigurable architecture (AVR) in FPGA for hardware applications is presented using a Genetic Programming Software on the development of an optimal reconfiguration for this AVR, in order to build a hardware capable of performing a given task in an embedded system. This proposal is a simple, flexible and efficient way to achieve appropriate applications in embedded systems, when compared to other reconfigurable hardware techniques. The representation of phenotype of the proposed evolutionary system is based on a bi-dimensional network function elements (EF). The GPLAB tool for MATLAB is used in Genetic Programming, and the solution found by this procedure is converted into a memory mapping to represent the best solution, where it is used to reconfigure the hardware. In the tests, GPLAB found results for logic circuits in a few generations, and for image filters containing efficient solutions, where there was little hardware occupation, especially memory, in the cases this has been presented, with a reduced chromosome size, shows a proposal efficiency. / O uso da tecnologia baseada em Field Programmable Gate Arrays (FPGAs), de forma reconfigurável, para a solução de diversos problemas atuais, tem se tornado um frequente objeto de estudo. Essa técnica é de aplicação viável e promissora na elaboração de sistemas embarcados, porém, a dificuldade em encontrar uma forma flexível e eficiente de realizar tal aplicação é o seu maior problema. Neste trabalho, é apresentada uma arquitetura virtual e reconfigurável (AVR) em FPGA para aplicações em hardware, utilizando um software de Programação Genética na elaboração de uma reconfiguração ótima para esta AVR, de forma a construir um hardware capaz de efetuar uma determinada tarefa em um sistema embarcado. Esta proposta é uma forma simples, flexível e eficiente de realizar aplicações adequadas em sistemas embarcados, quando comparada a outras técnicas de hardware reconfigurável. A representação do fenótipo no sistema evolutivo proposto se baseia em uma rede de elementos de função (EF) bidimensional. A ferramenta GPLAB, para MATLAB, é usada na Programação Genética, e a solução encontrada por esta é convertida em um mapeamento de memória com o cromossomo da melhor solução, onde este é usado para reconfigurar o hardware. Nos testes realizados, a GPLAB encontrou resultados para circuitos lógicos em poucas gerações, e para filtros de imagem encontrou soluções eficientes, onde ocorreu pouca ocupação de hardware, principalmente da memória nos casos apresentados, apresentando um cromossomo de tamanho reduzido, o que demonstra uma boa eficiência da proposta.

This diploma thesis deals with the use of Cartesian Genetic Programming (CGP) for combinational circuits design. The work addresses the issue of optimizaion of selected logic circuts, arithmetic adders and multipliers, using Cartesian Genetic Programming. The implementation of the CPG is performed in the Python programming language with the aid of NumPy, Numba and Pandas libraries. The method was tested on selected examples and the results were discussed.

Výzkum prezentovaný v této práci je zaměřen na evoluční optimalizaci vybraných komponent síťových aplikací určených pro monitorovací systémy vysokorychlostních sítí. Práce začíná studiem současných monitorovacích systémů. Jako experimentální platforma byl zvolen systém SDM (Software Defined Monitoring). Detailně bylo analyzováno zpracování síťového provozu, protože tvoří důležitou součást všech monitorovacích systémů. Jako demonstrační komponenty pro aplikaci optimálních technik navržených v této práci byly zvoleny klasifikátor aplikačních protokolů a hashovací funkce pro síťové toky. Evoluční algoritmy byly zkoumány s ohledem nejen na optimalizaci kvality zpracování dat danou síťovou komponentou, ale i na čas potřebný pro výpočet dané komponenty. Byly zkoumány jednokriteriální i vícekriteriální varianty evolučních algoritmů.     Byl navržen nový přístup ke klasifikaci aplikačních protokolů. Přesná i aproximativní verze klasifikátoru byla optimalizována pomocí CGP (Kartézské Genetické Programování). Bylo dosaženo výrazné redukce zdrojů a zpoždění v FPGA (Programovatelné Logické Pole) oproti neoptimalizované verzi. Speciální síťové hashovací funkce byly navrženy pomocí paralelní verze LGP (Lineární Genetické Programování). Tyto hashovací funkce vykazují lepší funkcionalitu oproti moderním hashovacím funkcím. S využitím vícekriteriální optimalizace byly vylepšeny výsledky původní jednokriteriální verze LGP. Paralelní zřetězené verze hashovacích funkcí byly implementovány v FPGA a vyhodnoceny za účelem hashování síťových toků. Nová rekonfigurovatelná hashovací funkce byla navržena jako kombinace vybraných hashovacích funkcí.  Velmi konkurenceschopná obecná hashovací funkce byla rovněž navržena pomocí multikriteriální verze LGP a její funkčnosti byla ověřena na reálných datových sadách v provedených studiích. Vícekriteriální přístup produkuje mírně lepší řešení než jednokriteriální LGP. Také se potvrdilo, že obecné implementace LGP a CGP jsou použitelné pro automatizovaný návrh a optimalizaci vybraných síťových komponent. Je však důležité zvládnout vícekriteriální povahu problému a urychlit časově kritické operace GP

When performing predictive data mining, the use of ensembles is claimed to virtually guarantee increased accuracy compared to the use of single models. Unfortunately, the problem of how to maximize ensemble accuracy is far from solved. In particular, the relationship between ensemble diversity and accuracy is not completely understood, making it hard to efficiently utilize diversity for ensemble creation. Furthermore, most high-accuracy predictive models are opaque, i.e. it is not possible for a human to follow and understand the logic behind a prediction. For some domains, this is unacceptable, since models need to be comprehensible. To obtain comprehensibility, accuracy is often sacrificed by using simpler but transparent models; a trade-off termed the accuracy vs. comprehensibility trade-off. With this trade-off in mind, several researchers have suggested rule extraction algorithms, where opaque models are transformed into comprehensible models, keeping an acceptable accuracy.In this thesis, two novel algorithms based on Genetic Programming are suggested. The first algorithm (GEMS) is used for ensemble creation, and the second (G-REX) is used for rule extraction from opaque models. The main property of GEMS is the ability to combine smaller ensembles and individual models in an almost arbitrary way. Moreover, GEMS can use base models of any kind and the optimization function is very flexible, easily permitting inclusion of, for instance, diversity measures. In the experimentation, GEMS obtained accuracies higher than both straightforward design choices and published results for Random Forests and AdaBoost. The key quality of G-REX is the inherent ability to explicitly control the accuracy vs. comprehensibility trade-off. Compared to the standard tree inducers C5.0 and CART, and some well-known rule extraction algorithms, rules extracted by G-REX are significantly more accurate and compact. Most importantly, G-REX is thoroughly evaluated and found to meet all relevant evaluation criteria for rule extraction algorithms, thus establishing G-REX as the algorithm to benchmark against. / <p>Avhandling framlagd 2007-06-01 vid Högskolan i Skövde.</p><p>Opponent: Rögnvaldsson, Thorsteinn, Professor, Sektionen för informationsvetenskap, Data- och Elektroteknik, Högskolan i Halmstad.</p>

Context Free Design Grammar (CFDG) is a programming language for defining recursive structures that can be used to create art. I use CFDG as a design space for genetic programming, experimenting with various options for crossover, mutation, and fitness. In this exploratory work, multiple generations are manually assessed to determine the usefulness of the mutation strategies and fitness functions. I find that simple value mutation and fitness that alters general program structure is not enough to produce an increase of interesting images in CFDG. I discuss these findings as well as future avenues of inquiry for genetic programming in artistic domains.

Submitted by Geyciane Santos (geyciane_thamires@hotmail.com) on 2015-06-22T14:42:52Z No. of bitstreams: 1 Tese- Patrícia Correia Saraiva.pdf: 5471120 bytes, checksum: faed3fa950294e70e5e4750ea26d9538 (MD5) / Approved for entry into archive by Divisão de Documentação/BC Biblioteca Central (ddbc@ufam.edu.br) on 2015-06-24T13:51:18Z (GMT) No. of bitstreams: 1 Tese- Patrícia Correia Saraiva.pdf: 5471120 bytes, checksum: faed3fa950294e70e5e4750ea26d9538 (MD5) / Approved for entry into archive by Divisão de Documentação/BC Biblioteca Central (ddbc@ufam.edu.br) on 2015-06-24T13:51:17Z (GMT) No. of bitstreams: 1 Tese- Patrícia Correia Saraiva.pdf: 5471120 bytes, checksum: faed3fa950294e70e5e4750ea26d9538 (MD5) / Approved for entry into archive by Divisão de Documentação/BC Biblioteca Central (ddbc@ufam.edu.br) on 2015-06-24T14:29:41Z (GMT) No. of bitstreams: 1 Tese- Patrícia Correia Saraiva.pdf: 5471120 bytes, checksum: faed3fa950294e70e5e4750ea26d9538 (MD5) / Made available in DSpace on 2015-06-24T14:29:41Z (GMT). No. of bitstreams: 1 Tese- Patrícia Correia Saraiva.pdf: 5471120 bytes, checksum: faed3fa950294e70e5e4750ea26d9538 (MD5) Previous issue date: 2014-02-28 / FAPEAM - Fundação de Amparo à Pesquisa do Estado do Amazonas / The volume of information encoded in the form of images has increased significantly in the last decades. Contributing to this scenario, the wide-spread use of mobile devices, such as tablets and smartphones, and even notebooks, which not only can take photos, but also easily send them to connected applications, such as web services and social networks. Nowadays, images are used in several applications, such as to record personal moments of people’s life or showing products in e-commerce online stores. As a consequence, not only does the volume of images increase, but also the interest in solutions able to retrieve these images. The main goal of this thesis is to investigate the impact of using genetic programming (GP) as a tool for combining different sources of evidence available when retrieving images. As case studies, we considered the application of GP in two different contexts: image retrieval on the Web using textual information automatically extracted from Web pages, and visual search by expanding the image query using information derived from different types of data, such as text and visual content. We evaluate the proposed expansion strategies in an application of visual search for products focused on e-commerce stores for the fashion domain. Experiments performed in the context of image retrieval on the Web showed that the evolutionary approach outperformed the best baseline with gains of 22.36% in terms of MAP. In the context of visual search for e-commerce applications, experimental results indicated that automatic expansion based on genetic programming is an effective alternative for improving the quality of image search results. When compared to a genetic programming system based only on visual information, the multimodal expansion achieved gains of at least 19% in all scenarios considered. When compared to a similar approach, but completely ad hoc, the GP framework achieved gains of up to 54% in terms of MAP. / O volume de informação codificada sob a forma de imagens tem aumentado de forma significativa nas últimas décadas. O uso cada vez mais frequente de tablets, smartphones, câmeras digitais e notebooks com suporte à aquisição de imagens e a facilidade para tornar essas imagens disponíveis publicamente em repositórios compartilhados, são fatores que contribuem ainda mais para este cenário. Atualmente, imagens são usadas nas mais diversas aplicações, seja para registrar momentos e ações em jornais e revistas eletrônicas, ou redes sociais, ou ainda para divulgar produtos em aplicações de comércio eletrônico. Na medida em que cresce o volume de imagens, cresce também o interesse por sistemas capazes de realizar busca em bases de dados de imagem. O objetivo principal desta tese é investigar o impacto do uso de programação genética (GP - Genetic Progamming) como ferramenta para combinar diferentes fontes de informação disponíveis durante a busca de imagens. Mais especificamente, foram abordados dois contextos distintos como estudos de caso: a busca de imagens na Web utilizando informação textual extraída automaticamente das páginas Web e, a busca visual por meio da expansão da imagem de consulta utilizando informação derivadas de diferentes modalidades de dados, como texto e conteúdo visual. Para avaliar as estratégias propostas para o contexto de busca visual, escolheu-se como estudo de caso a busca visual de produtos em lojas de comércio eletrônico voltadas para o segmento de moda. Os experimentos realizados no contexto de busca de imagens na Web mostraram que a abordagem evolucionária superou a melhor abordagem utilizada como baseline, com ganhos de 22,36% em termos de MAP. No cenário de busca visual de produtos em lojas de comércio eletrônico, os resultados experimentais mostraram que a expansão automática baseada em GP é uma alternativa efetiva para melhorar a qualidade dos resultados de um sistema de busca de imagens. Quando comparado a uma abordagem baseada somente em propriedades visuais, a expansão multimodal obteve ganhos de pelo menos 19% em todos os cenários de busca considerados. Quando comparado a uma abordagem similar, mas completamente ad hoc, o arcabouço baseado em GP obteve ganhos de até 54% em termos de MAP.

In modern computer systems, memory accesses and power management are the two major performance limiting factors. Accesses to main memory are very slow when compared to operations within a processor chip. Hardware write buffers, caches, out-of-order execution, and prefetch logic, are commonly used to reduce the time spent waiting for main memory accesses. Compiler loop interchange and data layout transformations also can help. Unfortunately, large data structures often have access patterns for which none of the standard approaches are useful. Using smaller data structures can significantly improve performance by allowing the data to reside in higher levels of the memory hierarchy. This dissertation proposes using lossy data compression technology called ’Compressive Hashing’ to create “surrogates”, that can augment original large data structures to yield faster typical data access. One way to optimize system performance for power consumption is to provide a predictive control of system-level energy use. This dissertation creates a novel instruction-level cost model called the variable-n-gram model, which is closely related to N-Gram analysis commonly used in computational linguistics. This model does not require direct knowledge of complex architectural details, and is capable of determining performance relationships between instructions from an execution trace. Experimental measurements are used to derive a context-sensitive model for performance of each type of instruction in the context of an N-instruction sequence. Dynamic runtime power prediction mechanisms often suffer from high overhead costs. To reduce the overhead, this dissertation encodes the static instruction-level predictions into a data structure and uses compressive hashing to provide on-demand runtime access to those predictions. Genetic programming is used to evolve compressive hash functions and performance analysis of applications shows that, runtime access overhead can be reduced by a factor of ~3x-9x.

Associative Memories (AMs) are essential building blocks for brain-like intelligent computing with applications in artificial vision, speech recognition, artificial intelligence, and robotics. Computations for such applications typically rely on spatial and temporal associations in the input patterns and need to be robust against noise and incomplete patterns. The conventional method for implementing AMs is through Artificial Neural Networks (ANNs). Improving the density of ANN based on conventional circuit elements poses a challenge as devices reach their physical scalability limits. Furthermore, stored information in AMs is vulnerable to destructive input signals. Novel nano-scale components, such as memristors, represent one solution to the density problem. Memristors are non-linear time-dependent circuit elements with an inherently small form factor. However, novel neuromorphic circuits typically use memristors to replace synapses in conventional ANN circuits. This sub-optimal use is primarily because there is no established design methodology to exploit the memristor's non-linear properties in a more encompassing way. The objective of this thesis is to explore denser and more robust AM designs using memristor networks. We hypothesize that such network AMs will be more area-efficient than the traditional ANN designs if we can use the memristor's non-linear property for spatial and time-dependent temporal association. We have built a comprehensive simulation framework that employs Genetic Programming (GP) to evolve AM circuits with memristors. The framework is based on the ParadisEO metaheuristics API and uses ngspice for the circuit evaluation. Our results show that we can evolve efficient memristor-based networks that have the potential to replace conventional ANNs used for AMs. We obtained AMs that a) can learn spatial and temporal correlation in the input patterns; b) optimize the trade-off between the size and the accuracy of the circuits; and c) are robust against destructive noise in the inputs. This robustness was achieved at the expense of additional components in the network. We have shown that automated circuit discovery is a promising tool for memristor-based circuits. Future work will focus on evolving circuits that can be used as a building block for more complicated intelligent computing architectures.

Chemical reaction networks are an unconventional computing medium that could benefit from the ability to form basic control systems. In this work, we demonstrate the functionality of a chemical control system by evaluating classic genetic algorithm problems: Koza's Santa Fe trail, Jefferson's John Muir trail, and three Santa Fe trail segments. Both Jefferson and Koza found that memory, such as a recurrent neural network or memories in a genetic program, are required to solve the task. Our approach presents the first instance of a chemical system acting as a control system. We propose a delay line connected with an artificial neural network in a chemical reaction network to determine the artificial ant's moves. We first search for the minimal required delay line size connected to a feed forward neural network in a chemical system. Our experiments show a delay line of length four is sufficient. Next, we used these findings to implement a chemical reaction network with a length four delay line and an artificial neural network. We use genetic algorithms to find an optimal set of weights for the artificial neural network. This chemical system is capable of consuming 100% of the food on a subset and greater than 44% of the food on Koza's Santa Fe trail. We also show the first implementation of a simulated chemical memory in two different models that can reliably capture and store information over time. The ability to store data over time gives rise to basic control systems that can perform more complex tasks. The integration of a memory storage unit and a control system in a chemistry has applications in biomedicine, like smart drug delivery. We show that we can successfully store the information over time and use it to act as a memory for a control system navigating an agent through a maze.