Search

Investigating the movements and site fidelity of individuals enhances our understanding of population ecology and structure. Killer whales occur around the coast of Iceland; however, information on the connectivity between different regions is limited to herring grounds, where they are observed frequently. In this study, we used photo-identification data to investigate the movements and site fidelity of whales within (South and West) and outside (Southwest, Northwest, Northeast and East) Icelandic herring grounds. Additionally, we used a 10-year photo-identification dataset in the South to investigate long-term site fidelity patterns to a single location. Of the 440 individuals sighted more than once, nearly half (48%) moved between herring grounds and site fidelity was higher within, compared to outside, herring grounds. Outside herring grounds, individuals showed: more movement from Southwest to West compare to South, indicating this region is not exclusively a passage between herring grounds; low site fidelity to the Northeast with fewer photographic matches to other regions, suggesting individuals found here may be part of an offshore population that occasionally visits the area; and low proportion of matches to other regions in the Northwest and East, although small sample sizes precluded firm conclusions. Finally, long-term residency of killer whales in the South showed dynamic patterns, likely caused by prey availability and/or environmental changes. This study elucidates the complexities of killer whale occurrence and connectivity within the North Atlantic and suggests population structure that should be further investigated for appropriate regional conservation assessments.

Sentiment analysis can be used for stock market prediction. However, existing research has not studied the impact of a user's financial background on sentiment-based forecasting of the stock market using artificial neural networks. In this work, a novel combination of neural networks is used for the assessment of sentiment-based stock market prediction, based on the financial background of the population that generated the sentiment. The state-of-the-art language processing model Bidirectional Encoder Representations from Transformers (BERT) is used to classify the sentiment and a Long-Short Term Memory (LSTM) model is used for time-series based stock market prediction. For evaluation, the Weibo social networking platform is used as a sentiment data collection source. Weibo users (and their comments respectively) are divided into Authorized Financial Advisor (AFA) and Unauthorized Financial Advisor (UFA) groups according to their background information, as collected by Weibo. The Hong Kong Hang Seng index is used to extract historical stock market change data. The results indicate that stock market prediction learned from the AFA group users is 39.67% more precise than that learned from the UFA group users and shows the highest accuracy (87%) when compared to existing approaches.

Medical imaging diagnosis increasingly relies on Machine Learning (ML) models. This is a task that is often hampered by severely imbalanced datasets, where positive cases can be quite rare. Their use is further compromised by their limited interpretability, which is becoming increasingly important. While post-hoc interpretability techniques such as SHAP and LIME have been used with some success on so-called black box models, the use of inherently understandable models makes such endeavors more fruitful. This paper addresses these issues by demonstrating how a relatively new synthetic data generation technique, STEM, can be used to produce data to train models produced by Grammatical Evolution (GE) that are inherently understandable. STEM is a recently introduced combination of the Synthetic Minority Oversampling Technique (SMOTE), Edited Nearest Neighbour (ENN), and Mixup; it has previously been successfully used to tackle both between class and within class imbalance issues. We test our technique on the Digital Database for Screening Mammography (DDSM) and the Wisconsin Breast Cancer (WBC) datasets and compare Area Under the Curve (AUC) results with an ensemble of the top three performing classifiers from a set of eight standard ML classifiers with varying degrees of interpretability. We demonstrate that the GE-derived models present the best AUC while still maintaining interpretable solutions.

Computational cost in metaheuristics such as Evolutionary Algorithms (EAs) is often a major concern, particularly with their ability to scale. In data-based training, traditional EAs typically use a significant portion, if not all, of the dataset for model training and fitness evaluation in each generation. This makes EAs suffer from high computational costs incurred during the fitness evaluation of the population, particularly when working with large datasets. To mitigate this issue, we propose a Machine Learning (ML)-driven Distance-based Selection (DBS) algorithm that reduces the fitness evaluation time by optimizing test cases. We test our algorithm by applying it to 24 benchmark problems from Symbolic Regression (SR) and digital circuit domains and then using Grammatical Evolution (GE) to train models using the reduced dataset. We use GE to test DBS on SR and produce a system flexible enough to test it on digital circuit problems further. The quality of the solutions is tested and compared against the conventional training method to measure the coverage of training data selected using DBS, i.e., how well the subset matches the statistical properties of the entire dataset. Moreover, the effect of optimized training data on run time and the effective size of the evolved solutions is analyzed. Experimental and statistical evaluations of the results show our method empowered GE to yield superior or comparable solutions to the baseline (using the full datasets) with smaller sizes and demonstrates computational efficiency in terms of speed.

Imbalanced datasets in medical imaging are characterized by skewed class proportions and scarcity of abnormal cases. When trained using such data, models tend to assign higher probabilities to normal cases, leading to biased performance. Common oversampling techniques such as SMOTE rely on local information and can introduce marginalization issues. This paper investigates the potential of using Mixup augmentation that combines two training examples along with their corresponding labels to generate new data points as a generic vicinal distribution. To this end, we propose STEM, which combines SMOTE-ENN and Mixup at the instance level. This integration enables us to effectively leverage the entire distribution of minority classes, thereby mitigating both between-class and within-class imbalances. We focus on the breast cancer problem, where imbalanced datasets are prevalent. The results demonstrate the effectiveness of STEM, which achieves AUC values of 0.96 and 0.99 in the Digital Database for Screening Mammography and Wisconsin Breast Cancer (Diagnostics) datasets, respectively. Moreover, this method shows promising potential when applied with an ensemble of machine learning (ML) classifiers., Comment: 7 pages, 4 figures, International Conference on Intelligent Computer Communication and Processing

Spacecraft are under constant threat from hypervelocity impacts by micrometeoroids and space debris. Due to their large surface area and hence increased probability of being impacted by debris, solar arrays are of particular concern. For larger, trackable debris objects, collision avoidance manoeuvres can be performed so as the minimise the damage caused to the spacecraft structure. However, for untrackable debris, particularly among the 1-3mm size range, strikes can occur at any stage and without any prior warning. During these collisions, momentum transfer takes place whereby the initial projectile momentum is imparted to the target. Due to the Momentum Enhancement Factor (MEF) caused by ejecta being expelled from the target, the actual value of transferred momentum can vary greatly depending on the product of initial projectile momentum and this factor. In this thesis, MEF values for a range of different impact conditions and scenarios are obtained. Hypervelocity impact simulations using smoothed particle hydrodynamics discretisation form the basis of these studies, enabling high impact velocities to be simulated which would otherwise be impossible using existing test equipment. Simplified models of Hubble Space Telescope solar arrays are implemented in order to gather large amounts of data and analyse the general trends and parameters which influence the MEF. On the whole it is shown that when the projectile momentum is corrected with the MEF to yield the actual momentum transferred, it can map to a whole range of potential projectile mass/velocity/shape combinations. This data alone is not enough to distinguish between impact types, but by reducing the very coarse MEF assumptions used in prior work, the uncertainties surrounding in situ perturbation data to validate debris environment models can begin to be reduced., Europe-Colorado Mobility Program, Aerospace Engineering

The deep sea is the largest ecosystem on Earth, yet little is known about the processes driving patterns of genetic diversity in its inhabitants. Here, we investigated the macro- and microevolutionary processes shaping genomic population structure and diversity in two poorly understood, globally distributed, deep-sea predators: Cuvier's beaked whale (Ziphius cavirostris) and Blainville's beaked whale (Mesoplodon densirostris). We used double-digest restriction associated DNA (ddRAD) and whole mitochondrial genome (mitogenome) sequencing to characterise genetic patterns using phylogenetic trees, cluster analysis, isolation-by-distance, genetic diversity and differentiation statistics. Single nucleotide polymorphisms (SNPs; Blainville's n = 43 samples, SNPs=13988; Cuvier's n = 123, SNPs= 30479) and mitogenomes (Blainville's n = 27; Cuvier's n = 35) revealed substantial hierarchical structure at a global scale. Both species display significant genetic structure between the Atlantic, Indo-Pacific and in Cuvier's, the Mediterranean Sea. Within major ocean basins, clear differentiation is found between genetic clusters on the east and west sides of the North Atlantic, and some distinct patterns of structure in the Indo-Pacific and Southern Hemisphere. We infer that macroevolutionary processes shaping patterns of genetic diversity include biogeographical barriers, highlighting the importance of such barriers even to highly mobile, deep-diving taxa. The barriers likely differ between the species due to their thermal tolerances and evolutionary histories. On a microevolutionary scale, it seems likely that the balance between resident populations displaying site fidelity, and transient individuals facilitating gene flow, shapes patterns of connectivity and genetic drift in beaked whales. Based on these results, we propose management units to facilitate improved conservation measures for these elusive species.

Accurate wavelength calibration is critical for qualitative and quantitative spectroscopic measurements. Many spectrometers for planetary exploration are equipped with onboard calibration sources. However, such calibration sources are not always available because planetary lander missions often have strong limitations in size and mass. In this study, we propose and validate a wavelength calibration method using solar Fraunhofer lines observed in reflective spectra. As a result, for a visible Raman spectrometer, the accuracy is better than 0.6 cm-1 in 0-4000 cm-1 range, and the magnesium abundance of olivine is estimated more accurately than 2%.

© The Authors, 2019. This article is distributed under the terms of the Creative Commons Attribution-Noncommercial-Share Alike 4.0 International License. The definitive version was published in Molecular Phylogenetics and Evolution (2019), doi:10.1016/j.ympev.2019.02.003., The advent of massive parallel sequencing technologies has resulted in an increase of studies based upon complete mitochondrial genome DNA sequences that revisit the taxonomic status within and among species. Spatially distinct monophyly in such mitogenomic genealogies, i.e., the sharing of a recent common ancestor among con-specific samples collected in the same region has been viewed as evidence for subspecies. Several recent studies in cetaceans have employed this criterion to suggest subsequent intraspecific taxonomic revisions. We reason that employing intra-specific, spatially distinct monophyly at non-recombining, clonally inherited genomes is an unsatisfactory criterion for defining subspecies based upon theoretical (genetic drift) and practical (sampling effort) arguments. This point was illustrated by a re-analysis of a global mitogenomic assessment of fin whales, Balaenoptera physalus spp., published by Archer et al. (2013), which proposed to further subdivide the Northern Hemisphere fin whale subspecies, B. p. physalus. The proposed revision was based upon the detection of spatially distinct monophyly among North Atlantic and North Pacific fin whales in a genealogy based upon complete mitochondrial genome DNA sequences. The extended analysis conducted in this study (1,676 mitochondrial control region, 162 complete mitochondrial genome DNA sequences and 20 microsatellite loci genotyped in 358 samples) revealed that the apparent monophyly among North Atlantic fin whales reported by Archer et al. (2013) to be due to low sample sizes. In conclusion, defining sub-species from monophyly (i.e., the absence of para- or polyphyly) can lead to erroneous conclusions due to relatively “trivial” aspects, such as sampling. Basic population genetic processes (i.e., genetic drift and migration) also affect the time to the most recent common ancestor and hence the probability that individuals in a sample are monophyletic., We are grateful to Hanne Jørgensen, Anna Sellas, Mary Beth Rew and Christina Færch-Jensen for technical assistance. We thank Drs. P. E. Rosel and K. D. Mullin (U.S. National Marine Fisheries Service Southeast Fisheries Science Center) and members of the U.S. Northeast and Southeast Region Marine Mammal Stranding Network and its response teams, including the International Fund for Animal Welfare, the Marine Mammal Stranding Center, Mystic Aquarium, the Riverhead Foundation for Marine Research and Preservation (K. Durham) and the Marine Mammal Stranding Program of the University of North Carolina Wilmington for access to fin whale samples from the western North Atlantic. We thank Gisli Vikingsson for providing samples. We are indebted to Dr. Eduardo Secchi for facilitating data sharing. Data collection in the Southern Ocean was conducted under research projects Baleias (CNPq grants 557064/2009-0 and 408096/2013-6), INTERBIOTA (CNPq 407889/2013-2) and INCT-APA (CNPq 574018/2008-5), of the Brazilian Antarctic Program and a contribution by the research consortium ‘Ecology and Conservation of Marine Megafauna – EcoMega-CNPq’. MAS was supported through a FCT Investigator contract funded by POPH, QREN European Social Fund, and Portuguese Ministry for Science and Education. Data collection in the Azores was funded by TRACE-PTDC/MAR/74071/2006 and MAPCET-M2.1.2/F/012/2011 [FEDER, COMPETE, QREN European Social Fund, and Proconvergencia Açores/EU Program]. Fin whale illustration herein is used with the permission of Frédérique Lucas. We acknowledge the Center for Information Technology of the University of Groningen for IT support and access to the Peregrine high performance-computing cluster.

The fin whale (Balaenoptera physalus) is the most abundant and widespread mysticete species in the Mediterranean Sea, found mostly in deep, offshore waters of the western and central portion of the region. In the Mediterranean, this species is known to feed mainly on krill, in contrast to its Atlantic counterpart, which displays a more diversified diet. The International Whaling Commission recognizes several managements units in the Atlantic and the Mediterranean Sea and the connectivity between these populations is still being debated. Questions remain about inter-individual feeding strategies and trophic ecology. The goal of this study was to compare isotopic niches of fin whales from the Mediterranean Sea and the Celtic Sea (North Atlantic). delta C-13 and delta N-15 values were analysed in 136 skin biopsies from free-ranging Mediterranean fin whales sampled in 2010 and 2011 during campaigns at sea. delta C-13 and delta N-15 values ranged from -20.4 to 17.1 parts per thousand and from -5.9 to 8.9 parts per thousand respectively. These values are in good agreement with those estimated previously from baleen plates from Mediterranean and North Atlantic fin whales. The narrow isotopic niche width of the Mediterranean fin whale (Standard Ellipses area SEAc) compared to the North Atlantic fin whale raises many concerns in the context of global changes and long-term consequences. One could indeed expect that species displaying narrow niches would be more susceptible to ecosystem fragmentation and other anthropogenic impacts.

Despite the limited research on cetaceans in Cabo Verde waters, 23 species have been documented (Hazevoet & Wenzel 2000, Hazevoet et al. 2010, Koenen et al. 2013, Berrow et al. 2015). The presence of several species is known only from strandings, such as the false killer whale, which has beached dead twice on Boavista in the 1980s–1990s, and again in 2007 (Hazevoet et al. 2010, Koenen et al. 2013). This is a CITES species, and Data Deficient according to IUCN (Taylor et al. 2008).

The importance of environmental fluctuations in the evolution of living organisms by natural selection has been widely noted by biologists and linked to many important characteristics of life such as modularity, plasticity, genotype size, mutation rate, learning, or epigenetic adaptations. In artificial-life simulations, however, environmental fluctuations are usually seen as a nuisance rather than an essential characteristic of evolution. HetCA is a heterogeneous cellular automata characterized by its ability to generate open-ended long-term evolution and “evolutionary progress”. In this paper, we propose to measure the impact of different types of environmental fluctuations in HetCA. Our results indicate that environmental changes induce mechanisms analogous to epigenetic adaptation or multilevel selection. This is particularly prevalent in two of the tested fluctuation schemes, which involve a round-robin inhibition of certain cell types, where phenotypic selection seems to occur., Funding for this work was provided by the Science Foundation Ireland and the ERC Advanced Grant EPNet #340828. Some of the simulations were run on the MareNostrum supercomputer of the Barcelona Supercomputing Center., Postprint (author's final draft)

During August-September 2014 and 2015, yachtbased surveys were conducted in the Cape Verde archipelago with the main objective of trying to locate humpback whales Megaptera novaeangliae during the late boreal summer (Berrow et al. 2014). Spring breeding humpbacks in Cape Verde waters are known to have their feeding grounds in the high Arctic (e.g. Wenzel et al. 2009). While these animals have generally left the breeding grounds by mid-May, a number of sightings during the summer months (June- August) have raised the possibility that animals from southern stocks may occasionally reach as far north as Cape Verde (Hazevoet et al. 2011).

The abundance of humpback whales occurring around Boa Vista, Cape Verde Islands, was estimated by markrecapture modelling. Photographs of markings on tail flukes of individual whales were collected during the breeding season. Sighting histories were constructed for each individual and an abundance of 171 animals was estimated using a Jolly-Seber mark-recapture model. Correcting for known biases arising due to sex-specific behaviour and temporary emigration insofar as possible, an estimate of 260 whales was obtained. This is significantly higher than the previous estimate of 99 humpback whales from this region. Due to limited survey effort it is not known how representative the study area is of the entire Cape Verde archipelago and this estimate may be considered to be biased low and serves as a minimum estimate. The high recapture probability (0.37) coupled with the low abundance is consistent with a small local population. The low survival rate (0.86) suggests possible emigration and further studies are needed to assess connectivity between humpback whales breeding in Cape Verde and other breeding locations. The amount of exchange between groups of whales breeding in Cape Verde and adjacent areas remains unknown. It is unclear whether the abundance estimate herein applies to part of an isolated population or part of a larger and continuous one., Apresenta-se uma estimativa da abundância de baleias-de-bossa na ilha da Boa Vista, Cabo Verde, realizada através de modelação por marcação e recaptura. Durante o período de reprodução, foram fotografadas marcas nas barbatanas caudais de indivíduos, tendo sido também construídos historiais de avistamentos. Aplicando o modelo de marcação e recaptura de Jolly-Seber chegou-se a uma estimativa de 171 baleias-de-bossa. Após correcção, tanto quanto possível, de desvios relacionados com comportamentos sexuais específicos e migração temporária, a estimativa foi alterada para 260 indivíduos. Este valor é consideravelmente maior do que a anterior estimativa para a mesma área, que identifica 99 indivíduos. Sendo os estudos ainda escassos, desconhece-se a representatividade da área de estudo face ao arquipélago de Cabo Verde no seu conjunto, pelo que esta estimativa deve estar enviesada negativamente e deve ser tomada como valor mínimo. A alta probabilidade de recaptura (0.37) e reduzida abundância são consistentes com uma população local e pequena. A baixa taxa de sobrevivência (0.86) sugere uma possível migração, pelo que devem ser realizados mais estudos no sentido de analisar a relação entre a reprodução das baleias-de-bossa em Cabo Verde e noutros locais. A frequência de mistura entre grupos de baleias que se reproduzem em Cabo Verde e em áreas adjacentes permanece desconhecida. Não é claro se a abundância aqui estimada se aplica a parte de uma população isolada ou a parte de uma população mais vasta e contínua.

This paper proposes a constrained-syntax genetic programming (GP) algorithm-for discovering classification rules in medical data sets. The proposed GP contains several syntactic constraints to be enforced by the system using a disjunctive normal form representation, so that individuals represent valid rule sets that are easy to interpret. The GP is compared with C4.5 in a real-world medical data set. This data set represents a difficult classification problem, and a new preprocessing method was devised for mining the data.

Artificial Immune Systems are computational algorithms which take their inspiration from the way in which natural immune systems learn to respond to attacks on an organism. This paper discusses how such a system can be used as an alternative to genetic algorithms as a way of exploring program-space in a system similar to genetic programming. Some experimental results are given for a symbolic regression problem. The paper ends with a discussion of future directions for the use of artificial immune systems in program induction.

For a given data set, its set of attributes defines its data space representation. The quality of a data space representation is one of the most important factors influencing the performance of a data mining algorithm. The attributes defining the data space can be inadequate, making it difficult to discover high-quality knowledge. In order to solve this problem, this paper proposes a Genetic Programming algorithm developed for attribute construction. This algorithm constructs new attributes out of the original attributes of the data set, performing an important preprocessing step for the subsequent application of a data mining algorithm.

Deriving Genetic Programming Fitness Properties by Static Analysis Colin G. Johnson The aim of this paper is to introduce the idea of using static analysis of computer programs as a way of measuring fitness in genetic programming. Such techniques extract information about the programs without explicitly running them, and in particular they infer properties which hold across the whole of the input space of a program. This can be applied to measure fitness, and has a number of advantages over measuring fitness by running members of the population on test cases. The most important advantage is that if a solution is found then it is possible to formally trust that solution to be correct across all inputs. This paper introduces these ideas, discusses various ways in which they could be applied, discusses the type of problems for which they are appropriate, and ends by giving a simple test example and some questions for future research.

peer-reviewed, Charles Darwin introduced the theory of evolution by natural selection in [Darwin and Wallace, 1858], and since that time the concept has remained unchanged at the highest level. However, there have been numerous and animated debates about its concrete implementation. Evolutionary Computation (EC) has taken inspiration from these debates, for example recently by reusing ideas from epigenetics [Tanev and Yuta, 2003]. We propose here to focus on a point which, although not central to the theory itself, comes up regularly in the polemics that have marked the eld of evolutionary biology: environmental uctuations. That is to say events uctuating randomly or regularly and modifying the optimal strategy maximizing an individual's tness. This thesis studies the e ects of environmental uctuations on natural selection in the context of computer simulations such as Genetic Programming (GP) and \open-ended" arti cial life simulations. The remainder of this chapter is organised as follows: Section 1.1 describes the motivation behind this research; Section 1.2 presents the research questions and the objectives addressed in the thesis; then Section 1.3 lists the contributions; nally Section 1.4 presents the structure of the thesis.

non-peer-reviewed, This thesis is concerned with leveraging the power of evolutionary computation to find structure or models within complex datasets. We opt for a data-centric approach to this issue, paying particular (but not exclusive) attention to real world datasets from the financial forecasting domain. Financial forecasting is an alluring endeavour that does not enjoy the sort of widely acknowledged success as do other popular areas of EC application such as circuit design, scheduling or industrial process control. Data believed to be pertinent to financial forecasting tends to be large and multi-dimensional, it is extremely non-linear and abundant with such a degree of noise that it has led to the popular school of thought that forecasting financial datasets is fundamentally impossible. Furthermore, there exists no widely accepted consensus, theoretical or otherwise, on the dynamics of financial markets. These and other troubling attributes make financial forecasting and the interpretation of financial data a fruitful testbed for evolutionary algorithms. By addressing the unique data challenges inspired by this domain we advance the understanding and capabilities of evolutionary computation. In particular our investigations result in the creation and proof of concept of two brand new evolutionary algorithms; the Hybrid Forecasting System -- used to evolve practical objectives to a problem, and, Evolutionary Multidimensional Scaling -- a new Multidimensional Scaling algorithm more appropriate to financial applications than existing MDS algorithms. In addition to these new algorithms this thesis demonstrates the application of EC to a new source of data (quantitative news sentiment), highlights a phenominon detrimental to EA's in noisy environments and demonstrates three new styles of data visualization.

peer-reviewed, The central hypothesis of this thesis is that it is possible to use a supervised machine learning technique, Genetic Programming (GP), to make accurate predictions and estimations regarding the endurance and retention of multi-level cell NAND Flash Memory devices. The retention of storage locations, or blocks, within these devices is the length of time for which they successfully retain their data, while their endurance is the number of times they can be programmed and erased prior to failure. Manufacturers currently place conservative speci cations on their devices since there is no technique available to quickly determine the actual endurance and retention capabilities of blocks within them. An extensive empirical evaluation of a number of MLC NAND Flash devices is completed, identifying features for use with GP, before expressions are evolved to make predictions and estimations regarding the retention and endurance of blocks. The empirical evaluation highlights the large variations in performance between blocks in di erent devices of the same speci cation, and even between blocks within the same device. As well as building a data set for later use with GP, the durations of program and erase operations are identi ed as features with which to make endurance predictions and estimations, while a relationship between block location and endurance is also established. GP is employed to evolve binary classi cation expressions, referred to as retention period classi ers, to predict whether blocks will correctly retain their data for a speci ed length of time. Following this, endurance classi ers are evolved to predict whether blocks will successfully complete a prede ned number of cycles. Finally, symbolic regression expressions are evolved, building on the earlier experiments, to estimate the actual number of cycles each block will complete prior to failure and are referred to as endurance estimators.

The archaeological site of Kharaneh IV in Jordan’s Azraq Basin, and its relatively near neighbour Jilat 6 show evidence of sustained occupation of substantial size through the Early to Middle Epipalaeolithic (c. 24,000 – 15,000 cal BP). Here we review the geomorphological evidence for the environmental setting in which Kharaneh IV was established. The on-site stratigraphy is clearly differentiated from surrounding sediments, marked visually as well as by higher magnetic susceptibility values. Dating and analysis of off-site sediments show that a significant wetland existed at the site prior to and during early site occupation (~ 23,000 – 19,000 BP). This may explain why such a substantial site existed at this location. This wetland dating to the Last Glacial Maximum also provides important information on the palaeoenvironments and potential palaeoclimatic scenarios for today’s 43 eastern Jordanian desert, from where such evidence is scarce.

peer-reviewed, The merit of Evolutionary Algorithms (EAs) as a means of automatic problem solving has been demonstrated numerous times on a diverse set of problem types across a range of different domains. The central hypothesis of this thesis is that by improving the expressiveness of EAs we can better support their deployment in domains in which context sensitive decision making is useful. After describing the principal structures and operations which allow EAs operate effectively as a general problem solving technique, we describe a sample problem and outline how two EA types, Genetic Programming (GP) and Grammatical Evolution (GE), might be configured to solve it. After some foundational elements of the discipline game design are presented, we highlight how a move towards more formal specifications of design elements presents new opportunities for the deployment of EAs as a means of Procedural Content Generation (PCG). Subsequently a set of experiments are described in which a system, designed to support encoding of data type information using a variant of GP called Strongly Typed Genetic Programming (STGP), is used to generate Player Character (PC) controllers for the digital video game Ms. Pac-Man. Following this an overview of Formal Grammars (FGs) is presented and the principal structures and operations of a third EA type, GE, are described. After which a number of more expressive FGs than Context Free Grammar (CFG), the grammar traditionally used with GE, are outlined. Finally, we outline a new GE variant designed to support usage Attribute Grammars (AGs), a means of specifying solution semantics in addition to syntax, and outline a set of experiments conducted using it. After highlighting the gains that can be made by using this GE variant in traditional problem domains such as symbolic regression, we discuss its potential as a means of PCG in digital video games. as a means of automatic problem solving has been demonstrated numerous times on a diverse set of problem types a

peer-reviewed, The merit of Evolutionary Algorithms (EAs) as a means of automatic problem solving has been demonstrated numerous times on a diverse set of problem types across a range of different domains. The central hypothesis of this thesis is that by improving the expressiveness of EAs we can better support their deployment in domains in which context sensitive decision making is useful. After describing the principal structures and operations which allow EAs operate effectively as a general problem solving technique, we describe a sample problem and outline how two EA types, Genetic Programming (GP) and Grammatical Evolution (GE), might be configured to solve it. After some foundational elements of the discipline game design are presented, we highlight how a move towards more formal specifications of design elements presents new opportunities for the deployment of EAs as a means of Procedural Content Generation (PCG). Subsequently a set of experiments are described in which a system, designed to support encoding of data type information using a variant of GP called Strongly Typed Genetic Programming (STGP), is used to generate Player Character (PC) controllers for the digital video game Ms. Pac-Man. Following this an overview of Formal Grammars (FGs) is presented and the principal structures and operations of a third EA type, GE, are described. After which a number of more expressive FGs than Context Free Grammar (CFG), the grammar traditionally used with GE, are outlined. Finally, we outline a new GE variant designed to support usage Attribute Grammars (AGs), a means of specifying solution semantics in addition to syntax, and outline a set of experiments conducted using it. After highlighting the gains that can be made by using this GE variant in traditional problem domains such as symbolic regression, we discuss its potential as a means of PCG in digital video games. as a means of automatic problem solving has been demonstrated numerous times on a diverse set of problem types a

peer-reviewed, The central hypothesis of this thesis is that it is possible to use a supervised machine learning technique, Genetic Programming (GP), to make accurate predictions and estimations regarding the endurance and retention of multi-level cell NAND Flash Memory devices. The retention of storage locations, or blocks, within these devices is the length of time for which they successfully retain their data, while their endurance is the number of times they can be programmed and erased prior to failure. Manufacturers currently place conservative speci cations on their devices since there is no technique available to quickly determine the actual endurance and retention capabilities of blocks within them. An extensive empirical evaluation of a number of MLC NAND Flash devices is completed, identifying features for use with GP, before expressions are evolved to make predictions and estimations regarding the retention and endurance of blocks. The empirical evaluation highlights the large variations in performance between blocks in di erent devices of the same speci cation, and even between blocks within the same device. As well as building a data set for later use with GP, the durations of program and erase operations are identi ed as features with which to make endurance predictions and estimations, while a relationship between block location and endurance is also established. GP is employed to evolve binary classi cation expressions, referred to as retention period classi ers, to predict whether blocks will correctly retain their data for a speci ed length of time. Following this, endurance classi ers are evolved to predict whether blocks will successfully complete a prede ned number of cycles. Finally, symbolic regression expressions are evolved, building on the earlier experiments, to estimate the actual number of cycles each block will complete prior to failure and are referred to as endurance estimators.

peer-reviewed, Multi-core processors are shared memory multiprocessors integrated on a single chip which offer significantly higher processing power than traditional, single core processors. However, as the number of cores available on a single processor increases, efficiently programming them becomes increasingly more complex, often to the point where the limiting factor in speeding up tasks is the software. This thesis presents Grammatical Automatic Parallel Programming (GAPP) which uses Grammatical Evolution to automatically generate natively parallel code on multi-core processors by directly embedding GAPP OpenMP parallelization directives in problem-specific Context Free Grammars. As a result, it obviates the need for programmers to think in a parallel manner while still letting them produce parallel code. We first perform a thorough analysis on the computational complexity of Grammatical Evolution using standard benchmark problems. This analysis results in an interesting experiment which produces a system capable of predicting on-the-fly the likelihood of a particular GE run being successful. A number of difficult proof of concept problems are examined in evaluating GAPP. The performance of the system on these informs the further optimization of both the design of grammars and fitness function to extract further parallelism. We demonstrate a surprising side effect of uncontrolled parallelism, which leads to the under-utilisation of the cores. This is addressed through the automatic generation of programs with controlled degree of parallelism. In this case, the automatically generated programs adapt to the number of cores on which they are scheduled to execute. Finally, GAPP is applied to Automatic Lockless Programming, an enormously difficult design problem, resulting in parallel code guaranteed to avoid locks on shared resources, thereby further optimizing the execution time. We then draw conclusions and make future recommendations on the use of evolutionary systems in the

non-peer-reviewed, This thesis is concerned with leveraging the power of evolutionary computation to find structure or models within complex datasets. We opt for a data-centric approach to this issue, paying particular (but not exclusive) attention to real world datasets from the financial forecasting domain. Financial forecasting is an alluring endeavour that does not enjoy the sort of widely acknowledged success as do other popular areas of EC application such as circuit design, scheduling or industrial process control. Data believed to be pertinent to financial forecasting tends to be large and multi-dimensional, it is extremely non-linear and abundant with such a degree of noise that it has led to the popular school of thought that forecasting financial datasets is fundamentally impossible. Furthermore, there exists no widely accepted consensus, theoretical or otherwise, on the dynamics of financial markets. These and other troubling attributes make financial forecasting and the interpretation of financial data a fruitful testbed for evolutionary algorithms. By addressing the unique data challenges inspired by this domain we advance the understanding and capabilities of evolutionary computation. In particular our investigations result in the creation and proof of concept of two brand new evolutionary algorithms; the Hybrid Forecasting System -- used to evolve practical objectives to a problem, and, Evolutionary Multidimensional Scaling -- a new Multidimensional Scaling algorithm more appropriate to financial applications than existing MDS algorithms. In addition to these new algorithms this thesis demonstrates the application of EC to a new source of data (quantitative news sentiment), highlights a phenominon detrimental to EA's in noisy environments and demonstrates three new styles of data visualization.

peer-reviewed, We present methods, observations and insights into the application of Evolu- tionary Algorithms(EA) to the problem of ash memory wear-out. In doing so we examine the union of two distinct cutting edge technologies: that of Non Volatile Memory(NVM) and that of EA, specifically the class of EAs known as Genetic Algorithms(GA). The complete adoption of ash memory for those applications that require non-volatile storage is inhibited by a small number of negative characteristics of ash, most notably wear-out and the data retention/endurance trade-off. This thesis describes how to build and validate an automated system that uses evolutionary search techniques to perform embodied evolution on hard silicon in order to find programming parameters that will reduce wear. We use the system to optimise the read, write and erase conditions of the device to enhance reliability. Since the exploration is done on actual silicon in real time, it is costly in both those terms. However, it provides a level of accuracy that could barely be approximated in simulation due to the complexity of the devices, the variance between storage elements and the sheer number of unknowns. We mitigate this cost with the use of small population methods and the structured inclusion of some acquired domain knowledge. Results are calculated on a per device basis, with derived solutions com- pared to baseline results for that device. They demonstrate an increase in endurance of up to 300% per device. A blueprint for future experimentation with sequential access, or NAND ash memory, is presented. Although there has been an embargo placed on this thesis, with the result that very little can be published from it, this work has had considerable impact, getting coverage in international popular science publications, as well as leading to research funding involving several institutions and companies.

peer-reviewed, Charles Darwin introduced the theory of evolution by natural selection in [Darwin and Wallace, 1858], and since that time the concept has remained unchanged at the highest level. However, there have been numerous and animated debates about its concrete implementation. Evolutionary Computation (EC) has taken inspiration from these debates, for example recently by reusing ideas from epigenetics [Tanev and Yuta, 2003]. We propose here to focus on a point which, although not central to the theory itself, comes up regularly in the polemics that have marked the eld of evolutionary biology: environmental uctuations. That is to say events uctuating randomly or regularly and modifying the optimal strategy maximizing an individual's tness. This thesis studies the e ects of environmental uctuations on natural selection in the context of computer simulations such as Genetic Programming (GP) and \open-ended" arti cial life simulations. The remainder of this chapter is organised as follows: Section 1.1 describes the motivation behind this research; Section 1.2 presents the research questions and the objectives addressed in the thesis; then Section 1.3 lists the contributions; nally Section 1.4 presents the structure of the thesis.

peer-reviewed, Multi-core processors are shared memory multiprocessors integrated on a single chip which offer significantly higher processing power than traditional, single core processors. However, as the number of cores available on a single processor increases, efficiently programming them becomes increasingly more complex, often to the point where the limiting factor in speeding up tasks is the software. This thesis presents Grammatical Automatic Parallel Programming (GAPP) which uses Grammatical Evolution to automatically generate natively parallel code on multi-core processors by directly embedding GAPP OpenMP parallelization directives in problem-specific Context Free Grammars. As a result, it obviates the need for programmers to think in a parallel manner while still letting them produce parallel code. We first perform a thorough analysis on the computational complexity of Grammatical Evolution using standard benchmark problems. This analysis results in an interesting experiment which produces a system capable of predicting on-the-fly the likelihood of a particular GE run being successful. A number of difficult proof of concept problems are examined in evaluating GAPP. The performance of the system on these informs the further optimization of both the design of grammars and fitness function to extract further parallelism. We demonstrate a surprising side effect of uncontrolled parallelism, which leads to the under-utilisation of the cores. This is addressed through the automatic generation of programs with controlled degree of parallelism. In this case, the automatically generated programs adapt to the number of cores on which they are scheduled to execute. Finally, GAPP is applied to Automatic Lockless Programming, an enormously difficult design problem, resulting in parallel code guaranteed to avoid locks on shared resources, thereby further optimizing the execution time. We then draw conclusions and make future recommendations on the use of evolutionary systems in the

The archaeological site of Kharaneh IV in Jordan’s Azraq Basin, and its relatively near neighbour Jilat 6 show evidence of sustained occupation of substantial size through the Early to Middle Epipalaeolithic (c. 24,000 – 15,000 cal BP). Here we review the geomorphological evidence for the environmental setting in which Kharaneh IV was established. The on-site stratigraphy is clearly differentiated from surrounding sediments, marked visually as well as by higher magnetic susceptibility values. Dating and analysis of off-site sediments show that a significant wetland existed at the site prior to and during early site occupation (~ 23,000 – 19,000 BP). This may explain why such a substantial site existed at this location. This wetland dating to the Last Glacial Maximum also provides important information on the palaeoenvironments and potential palaeoclimatic scenarios for today’s 43 eastern Jordanian desert, from where such evidence is scarce.

The archaeological site of Kharaneh IV in Jordan’s Azraq Basin, and its relatively near neighbour Jilat 6 show evidence of sustained occupation of substantial size through the Early to Middle Epipalaeolithic (c. 24,000 – 15,000 cal BP). Here we review the geomorphological evidence for the environmental setting in which Kharaneh IV was established. The on-site stratigraphy is clearly differentiated from surrounding sediments, marked visually as well as by higher magnetic susceptibility values. Dating and analysis of off-site sediments show that a significant wetland existed at the site prior to and during early site occupation (~ 23,000 – 19,000 BP). This may explain why such a substantial site existed at this location. This wetland dating to the Last Glacial Maximum also provides important information on the palaeoenvironments and potential palaeoclimatic scenarios for today’s 43 eastern Jordanian desert, from where such evidence is scarce.

The archaeological site of Kharaneh IV in Jordan’s Azraq Basin, and its relatively near neighbour Jilat 6 show evidence of sustained occupation of substantial size through the Early to Middle Epipalaeolithic (c. 24,000 – 15,000 cal BP). Here we review the geomorphological evidence for the environmental setting in which Kharaneh IV was established. The on-site stratigraphy is clearly differentiated from surrounding sediments, marked visually as well as by higher magnetic susceptibility values. Dating and analysis of off-site sediments show that a significant wetland existed at the site prior to and during early site occupation (~ 23,000 – 19,000 BP). This may explain why such a substantial site existed at this location. This wetland dating to the Last Glacial Maximum also provides important information on the palaeoenvironments and potential palaeoclimatic scenarios for today’s 43 eastern Jordanian desert, from where such evidence is scarce.