An essential component of the game is the box it comes in. Though it seems like a small object, it is the first door to the world it encapsulates in itself. Box makes an impression, and perceptions are associated with it. This paper investigates board game's box art and related perceptions. A disappointing box art results in the game being left un-purchased on the store shelves. This is not the only consequence. An undesirable impression from the box art remains with the player, although they enjoy the game. While the games have the potential to convey intended benefits, the box is the only chance for game designers to evoke buyers' attention, if the game has not been advertised by other means. It is significant to understand how players associate with box art, their perceptions, and their desirability. This paper has investigated box art design for its desirability, the first impression it conveys, and perception of it. We present a series of questions that help test the box art of board games. The questions are broken down into categories to investigate different aspects of the box art. The participants have also drawn the box art for a particular game setting that informs about their preferred design.

This project presents an evolutionary algorithm for texture generation that allows users to choose and manipulate camouflage patterns. The initial results of a pilot study provide some insight into usability and the users’ ability to replicate a target pattern. The result is an evaluation of gathered data showing user tendencies and how they engage with the system. These tendencies include significantly different completion times for target patterns varying in complexity. Additionally, participants mostly agreed that the tool is helpful for future games and objects other than camouflage skins. The findings suggest potential applications for artificial intelligence in enhancing user customization and design flexibility. Further research must address technical limitations and explore broader game industry implications.

Camouflage serves a dual purpose in games, imparting believability of the world and narrative and being a game asset used by players for customization. Using the U.S. Woodland Battle Dress Uniform or M81 as an inspiration for the pattern, we present a system for generating camouflage using a process based on present military evaluations. The proposed approach is a genetic algorithm that uses image processing/analysis techniques to assess the generated texture and can create camouflage that allows the soldier model to blend with the environment. A human evaluation was conducted and shows that the fitness metric used by the system is a suitable surrogate.

A well-designed video game soundtrack can significantly affect human game perception, especially when there is an intuitive link between musical and game features. The soundtrack intuitiveness can be increased by making it adaptive and dependent on players’ actions. The tempo is one of the music characteristics, and this change is relatively easy to distinguish even for non-musicians because it is often interpreted as a speed. This work examines the existence of different correlations between players’ in-game actions and soundtrack tempo. Authors suppose that results of conducted playtesting with humans can improve game development from the musical side, increasing players’ engagement with the game. The playtesting is done based on a simple runner game called MAK, which was developed for scientific purposes. This research aims to find intuitive dependencies between six considered game actions and soundtrack tempo.

Contact networks are used as a representation for the modeling of illness transmission. In this study, we represent not only the links of the transmissions but also utilize a weighted graph to represent the probability of transfer. These graphs can be large and complex when taking into account the number of contacts used in tracing. Compression of the graph allows for the development of community detection as well as providing a simpler graph. By examining the contact networks developed by an evolutionary algorithm for compression, it is discovered that the choice of fitness function and the appropriate weighting of edges leads to a different compressed graph, finding different connected communities; this is also true when compared to the communities identified by the Louvain community detection algorithm. This demonstrates the importance of considering weighting in contact networks, and suggests that in the future an understanding of the community structure should be utilized by public health officials.

This study compares a current representation for evolving networks to model epidemic spread with a novel representation also studied in a companion paper. This study applies a powerful diversity-friendly algorithm called ring optimization to this novel representation. The problem addressed is that the baseline method is found to optimize only locally; use of the novel representation improves the situation, but not much. The use of ring optimization yields similar or better performance for the ability of the evolved networks to model epidemics while substantially increasing the diversity of those networks.

This study introduces a new game that models competition in foraging behavior. Two moose decide, in each time period, which of three foraging areas to visit. Moose in the same foraging area fight, gaining no forage and also damaging some forage during their conflict. Moose alone in a foraging area eat, with the forage in each field being replenished with a logistic growth model. This creates a relatively complex game with a rich strategy space in which the moose try to maximize their forage intake. The game is a coordination game, as the moose try to avoid conflict which does not maximize forage intake. The paper reports the results of two student competitions at Innopolis University and performs agent evolution to verify the existence of a rich strategy space for the game.

Cities, while exciting in their visualization and permitting several layouts, do not take into account the placement of crucial characters which might be part of the narrative. Narrative graphs, a connected graph of all potential and existing relations within a game, can enable an ability to find a Nonplayer Character (NPC) who is likely to live nearby, under the assumption that those who interact most frequently are also close in distance. We examine the use of an evolutionary graph compression method and a method using simulated diffusion to cluster features based on relational information about players to generate relationally intimate groups. This clustering can be used to generate information about the game world and cities to inform PCG as to how the connectivity of these areas is, and should be, arranged. The algorithms are validated as being human competitive.

Evolutionary Algorithms (EA) capacitate a myriad of possibilities and creativity. Therefore, while teaching evolutionary programming to students of Computer Science, we decided to equip them with the technical details and then let them explore the creative aspects themselves. As part of the introductory course on Artificial Intelligence, an evolutionary generation of art is an assignment The teaching goals were to inform students of creative aspects of EA as well as invoke critical thinking and analysis in them. The students generated images via EA, and the only restriction was on the pixel size for input. The assignment was well-received, despite its complexity. Students engaged with the art generation and felt confident in expressing their perspectives of creativity and art The assignment evaluation as a contest allowed for judges from diverse domains such as Computer scientists and artists. The goals of the course are met as the students were trained not only in technical details but also realized the capacity and generative power of algorithms, as they practically simulated and expressed their creative thoughts via this contest Furthermore, the students engaged with the philosophical foundations of computational creativity effectively by the design process, and a report completed along with the work required them to develop their own artistic exegesis of the algorithm.

Parkinson's disease is a neurodegenerative disease that affects close to 10 million with various symptoms including tremors and changes in gait. Observing differences or changes in an individual's manifestations of gait may provide a mechanism to identify Parkinson's disease and understand specific changes. In this study, timeseries data from both Control subjects and Parkinson's disease patients was modelled with symbolic regression and extreme gradient boosting. Model effectiveness was analyzed along with the differences in the models between modelling strategies, between Control subjects and Parkinson's disease patients, and between normal walking and walking while under a cognitive load. Both modelling strategies were found to effective. The symbolic regression models were more easily interpreted, while extreme gradient boosting had higher overall accuracy. Interpretation of the models identified certain characteristics that distinguished Control subjects from Parkinson's disease patients and normal walking conditions from walking while under a cognitive load.

Epidemic contact tracing examines the movement of infection through a population based upon links in a contact network, and weighted networks represent the potential of transfer of the contagion. Graph compression reduces the size of a network by merging groups of nodes into supernodes. This study considers the use of genetic algorithms to select the nodes to be merged, grouping together highly connected sections of the graphs. Examined is a dataset that is extracted from contacts that occurred during several days of the "Infectious: Stay Away" event. The incorporation of weights, to indicate the strength of interactions between individuals, is an important contribution of this work. The demonstrated outcomes are that by including weighted information on the edges, there is more effective detection of highly interacting subgroups when compared to the unweighted version of graphs. These methods not only compress the networks with a low rate of distortion, but also the identification of supernodes in the networks allows for better targeting of interventions by public health upon individuals in such groups. This is crucial because when one member becomes infected, all members of the group are exposed to the contagion.

Parkinson's Disease is a disorder with diagnostic symptoms that include a change to a walking gait. The disease is problematic to diagnose. An objective method of monitoring the gait of a patient is required to ensure the effectiveness of diagnosis and treatments. We examine the suitability of Extreme Gradient Boosting (XGBoost) and Artificial Neural Network (ANN) Models compared to Symbolic Regression (SR) using genetic programming that was demonstrated to be successful in previous works on gait. The XGBoost and ANN models are found to out-perform SR, but the SR model is more human explainable.

Smart devices are becoming more universally adopted and can be used to track and model user activity and monitor for abnormalities. Deviations from what is expected may indicate that a fall is imminent or that an injury has been sustained. Healthcare practitioners can use descriptive models of human kinematics as a tool to monitor patient recovery. This work extends previous work which generated descriptive nonlinear symbolic models of human kinematics with genetic programming. Previously, linear models were developed and compared to the nonlinear models. Although the linear models fit the data well, they were significantly worse than the nonlinear models. In this phase of the project, ensembles of nonlinear models were created to more accurately fit and classify data. Different model selection strategies for the ensembles were investigated. As one would expect, ensembles of models were significantly better than a single model classifier. It was also observed that, although more models in the ensemble yielded better results, only 2 models were required to obtain significantly better results. It was also observed that a random model selection strategy for the ensembles produced competitive results when compared to a more rigorous model selection strategy.

Network graphs appear in a number of important biological data problems, recording information relating to protein-protein interactions, gene regulation, transcription regulation and much more. These graphs are of such a significant size that they are impossible for a human to understand. Furthermore, the ever-expanding quantity of such information means that there are storage issues. To help address these issues, it is common for applications to compress nodes to form supernodes of similarly connected components. In previous graph compression studies it was noted that such supernodes often contain points from disparate parts of the graph. This study aims to correct this flaw by only allowing merges to occur within a local neighbourhood rather than across the entire graph. This restriction was found to not only produce more meaningful compressions, but also to reduce the overall distortion created by the compression for two out of three biological networks studied.

Parkinson's disease (PD) is a degenerative disorder of the central nervous system that has many debilitating symptoms which affect the patient's motor system and can cause significant changes in their gait. By using genetic programming, we aim to develop descriptive symbolic nonlinear models of PD patient gait from time series data recorded from pressure sensors under subjects' feet. When compared to popular types of linear regression (OLS and LASSO), the nonlinear models fit their data better and generalize to unseen data significantly better. It was found that models developed for healthy control subjects generalized to other control subjects well, however the models trained on subjects with PD did not generalize well to other PD patients, which complicates the issue of being able to detect the progression of the disease. It is suspected that health care professionals can have difficulty classifying PD due to a lack of accurate data from patient reports; having individually trained models for active monitoring of patients would help in effectively diagnosing PD.

The problem of smoking is an extremely relevant issue. There is a need for reducing smoking among the population, especially children. The aim is to develop an automated solution for cigarette censoring. The application of the automated censoring of cigarettes in images or videos may have a positive effect on the problem of reducing smoking among the population, also minimizing the risk of a child being affected by harmful information. In the process of creation of the solution, various segmentation models and loss functions were tested. The solution is based on the DeepLab architecture for image semantic segmentation, which extracts the cigarette pixels from the image. These pixels are then transformed, so the cigarette in the image becomes unidentifiable. Achieved results are considered to be satisfactory, removing cigarettes with minimal disruption to the image. Mean Intersection over Union of 0.791 shown by DeepLab in the segmentation task and pixel swapping method simplicity are combined in the overall good performance.

Colours and contrast are significant for aesthetics and for readability reference leading to a need to identify the correlation between the player preference of colours in general and to game objects. The object chosen for the experiment is dice. Dice come in a variety of colours and designs and with its simple usability mechanics, is a compelling object for investigation. These dice are examined for the correlation to players colour preferences and a set of contrasting dice are examined for their readability errors. It was found that the die with minimal contrast provides for more readability errors and greater time required to understand the roll. Furthermore, it has been identified that dice colour preference does not correlate with colour preference.

Tabletop role-playing Games (RPG), such as Dungeons & Dragons (D&D) use dice in order to control the outcome actions by characters when the Game Master needs to introduce randomness. While dice are fundamental to such games, the examination of dice as objects of design has not been explored. This study examines fifty-nine participants (thirty familiar with the D20 set system) and asks them to examine two 7-die sets commonly used in D&D, the first set being a common set of polyhedrons, and the other set designed to replicate the objects used by a Wizard. It examines the fairness perceptions of the participants and finds that players who have experience with the polyhedral set in the past are more likely to accept the fairness of Wizard dice, and that all players are more likely to accept the fairness of the Wizard set after a play session.

The fertility of two structures in an evolutionary computation system is the expected fitness of their potential offspring. The study uses the problem of locating interesting fractals to introduce an application of fertility intended to reduce user fatigue in Interactive Evolutionary Computation with a human-in-the-loop evaluation method. High fertility sets of fractal parameters are shown to substantially increase the performance of evolution using small population size, a surrogate for human-driven selection. The high fertility sets of fractal parameters are a form of automatically generated content that is part of an application intended to permit a user to find pleasing fractals.

Side Effect Machines (SEMs) are an extension of finite state machines which place a counter on each node that is incremented when that node is visited. Previous studies examined a genetic algorithm to discover node connections in SEMs for edit metric decoding for biological applications, namely to handle sequencing errors. Edit metric codes, while useful for decoding such biologically created errors, have a structure which significantly differentiates them from other codes based on Hamming distance. Further, the inclusion of biologically- motivated restrictions on allowed words makes development of decoders a bespoke process based on the exact code used. This study examines the use of evolutionary programming for the creation of such decoders, thus allowing for the number of states to be evolved directly, not witnessed in previous approaches which used genetic algorithms. Both direct and fuzzy decoding are used, obtaining correct decoding rates of up to 95% in some SEMs.

Smart devices are becoming an extension of ourselves that contain sensitive information and are often targeted for theft. The development of an intelligent and reliable means of user identification and authentication is critical. Not only can the development of user models performing tasks be used for user and task identification, but systems can also notify individuals if there is a potential health concern. The construction of an idealized model of human locomotion may give medical care providers a better understanding of individual differences and guide therapy and treatment. Data was gathered from a smartwatch worn by six subjects performing five different tasks and Genetic Programming was used to perform symbolic regression - a model free, nonlinear type of regression analysis. Symbolic regression was applied to smartwatch data and a collection of nonlinear closed form symbolic mathematical models were generated. Not only did these models fit the data well, but they provided insight into the underlying system. With only 5 seconds of unseen data, the models could classify which subjects were performing which task with 83.9% accuracy when chance was only 3.33%.

The software is often produced under significant time constraints. Our idea is to understand the effects of various software development practices on the performance of developers working in stressful environments, and identify the best operating conditions for software developed under stressful conditions collecting data through questionnaires, non-invasive software measurement tools that can collect measurable data about software engineers and the software they develop, without intervening their activities, and biophysical sensors and then try to recreated also in different processes or key development practices such conditions.

Dice since the classical era have been some of the most popular object in games, with examples of six sized dice as we know them found in Egyptian grave goods and Roman burials. Over more than a millennium, they have seen diversification. How people have adjusted to, affected by, or embraced these modifications is a subject of this paper. The study presented investigates people's perception of fairness in a pair of dice with a 2D6 distribution. Three pair of dice were presented, each modeling the distribution and proven mathematically to be fair, to the participants and they were asked about their opinion regarding the fairness of these dice. Participants were then able to use the dice via playing the game snakes and ladders. The results so far suggests variety of opinions regarding people's perception of fairness. The study also revealed that participants sometimes preferred to play with the dice they did not consider fair because of the unusual design of the dice or because in the dice rolls, they were getting higher numbers as outcomes.

Storage and processing of biological networks is challenging and costly due to the large sizes of many of these networks. Compression of such graphs is one possible solution to this problem. This study presents two single-objective genetic algorithms, along with one multi-objective algorithm, to address the problem of graph compression. The fitness functions were both based on the concept of merging nodes based on “similarity” but each defined that similarity in a different way. The multiobjective GA based on NSGA-II worked to find a balance between the compression ratio and the similarity. The methods were applied to three different biological networks with different characteristics. The single-objective GAs were first applied to these networks for a fixed compression ratio. Then based on the results of the multiobjective GA, target compression ratios were chosen for each graph and the single-objective GAs were applied to this target. Applying the single-objective GAs to a target identified in this manner was significantly more successful than using the results from the multiobjective GA for the same compression ratio.

Partitioning regression is the simultaneous fitting of multiple models to a set of data and partitioning of that data into easily modelled classes. The key to partitioning regression with evolution is minimum error assignment during fitness evaluation. Assigning a point to the model for which it has the least error while using evolution to minimize total model error encourages the evolution of models that cleanly partition data. This study demonstrates the efficacy of partitioning regression using two or three models on simple bivariate data sets. Two novel multi-model representations, a simple evolutionary parameters setting algorithm and one based on using directed acyclic graphs representations for genetic programming are used. Possible generalizations to the general case of clustering are outlined.

The idea of using the gait of a walking person as a biometric identification method has been seen in a number of proposed authentication methods, yet previous works focus on the addition of other authentication methods along with the gait, or require a stationary sensor attached to the hip of the user. This paper uses Genetic Programming to model an identification gait fingerprint for two users, whose walking data was recorded from the accelerometer in a commercially available phone. With the phone freely placed within a pocket, users moved without a fixed protocol at a normal, nonuniform pace. This design of data collection more closely matches the real world applications of such a method. The highly specialized Genetic Programming system with multiple modular enhancements was implemented to perform symbolic regression. The system was demonstrated to be robust to noise and was able to effectively model each dataset with high accuracy. It was also determined that a model could be generated for a subject's whole dataset from only a single step's worth of data. Top models were applied to other subject's data in order to evaluate the uniqueness of these mathematical models.

Many real-world graphs, including those storing various forms of biological data, are of such large size that storing and processing their information has too high a cost. As a result, one possible solution is to compress the graphs by merging nodes into supernodes. This study introduces a genetic algorithm for graph compression that is based on the similarity of nodes, where two nodes are considered similar if a high proportion of their neighbours are in common. The methodology was applied to three real-world graphs storing widely varying data, as well as the gene regulatory network of E. coli. This study used a fixed compression rate of 25% as a target for the graphs. Results for a parameter study of variation operators exhibit a strong preference for crossover, in comparison to mutation which was found to be disruptive to graph structure.

Smartphones have ubiquitously integrated into our home and work environments. It is now a common practice for people to store their sensitive and confidential information on their phones. This has made it extremely important to authenticate legitimate users of a phone and block imposters. In this paper, we demonstrate that the motion dynamics of smartphones, captured using their built in accelerometers, can be used for accurate user identification. We call this mechanism gait fingerprinting. To this end, we first collected the acceleration data from multiple users as they walked with a smartphone placed freely in their pants pockets. Next, we studied the application of different feature extraction, feature selection and classification techniques from the machine learning literature on these data. Through extensive experimentation, demonstrated is that simple time domain features extracted from these data, which are further optimized using stepwise linear discrimination analysis, can be used to train artificial neural networks to identify legitimate user and block imposter with an average accuracy of 95%.

Side Effect Machines (SEMs) have been used as a evolutionary representation in a variety of studies dealing with the classification of data for bioinformatic studies. However, up to this point there has been no formalism of the SEM in terms of its representational ability and placement within the Chomsky hierarchy; only a statement that it is a generalization of a Deterministic Finite Automation (DFA), without proof, has been provided. This paper aims to rectify that situation by presenting a formal look at SEMs in terms of the languages which they are known to accept. We give a constructive proof of how a SEM is a generalization of a DFA and are therefore able to be used to accept languages. Constructive proofs for SEMs accepting families of context-free and context-sensitive languages are also provided.

Agents for playing iterated prisoner's dilemma are commonly trained using a coevolutionary system in which a player's score against a selection of other members of an evolving population forms the fitness function. In this study we examine instead a version of evolutionary iterated prisoner's dilemma in which an agent's fitness is measured as the average score it obtains against a fixed panel of opponents called an examination board. The performance of agents trained using examination boards is compared against agents trained in the usual coevolutionary fashion. This includes assessing the relative competitive ability of players evolved with evolution and coevolution. The difficulty of several experimental boards as optimization problems is compared. A number of new types of strategies are introduced. These include sugar strategies which can be exploited with some difficulty and treasure hunt strategies which have multiple trapping states with different levels of exploitability. The degree to which strategies trained with different examination boards produce different agents is investigated using fingerprints.

Quaternary error-correcting codes defined over the edit metric may be used as labels to track the origin of sequence data. When used in such applications there are typically additional restrictions that are biologically motivated, such as a required GC content or the avoidance of certain patterns. As a result such codes can not be expected to have a regular structure, making decoding particularly challenging. Previous work on decoding edit codes considered the use of side effect machines for decoding, successfully decoding up to 93.86% of error vectors. In this study the recentering/restarting algorithm is used in combination with side effect machines and an alternative representation based upon transpositions. Using the same data as in the previous work, the rate of successful decoding was significantly improved, with many cases obtaining rates very close to 100%.

The Multiple Worlds Model of evolution is a spatially structured evolutionary algorithm which uses the ideas of Darwin's finches as a motivating idea. Through multiple populations separated genetically but with a unified fitness evaluation, Multiple Worlds acts to partition data via specialization. Each of the populations must specialize in order to gain fitness, or a population can be reduced to little or no fitness. Such a drop in fitness implies that the number of populations, each representing a class in the data, is too large. The number of natural classes in the data is discovered by the algorithm via an analog to biological extinction. This study examines the application of this method to discovery of degenerate motifs on two types of data. The first is the classification of synthetic motifs, created by a self-driving finite state machine, selected to yield high-entropy data. The second is a biological example comprising two classes of data drawn from a Human Leukocyte Antigen data set. The classifiers found not only allow for the division of the data, but are expressed as degenerate motifs; granting researchers a comprehensible, reusable result.

In this study we look at a novel evolutionary technique known as the multiple worlds model for unsupervised classification of sequences via evolved motifs. This evolutionary algorithm uses the biological inspirations of species and species extinction as features modeled in the evolution in order to provide classifiers where the number of classes is not known a priori. In the multiple worlds model a number of populations which do not interbreed compete in fitness evaluation. Sequence motifs are small, biologically significant DNA/RNA or amino acid segments. They are represented as strings of symbols and wild cards. The model works well to locate classification motifs for sequences whose classes have statistical deviations such as those with a GC content difference or those created from differing Self-Driving Markov models. The creation of such classifiers will allow biologists to examine large sets of sequences in order to discover significant features.

The Mandelbrot set is a famous fractal. It serves as the source of a large number of complex mathematical images. Evolutionary computation can be used to search the Mandelbrot set for interesting views. This study compares the results of using several different fitness functions for this search. Some of the fitness functions give substantial control over the appearance of the resulting views while others simply locate parts of the Mandelbrot set in which there are complicated structures. All of the fitness functions are based on finding desirable patterns in the number of iterations of the basic Mandelbrot formula to diverge on a set of points arranged in a regular grid near the boundary of the set. It is shown that using different fitness functions causes an evolutionary algorithm to locate difference types of views into the Mandelbrot set.

We present a novel method for the creation of photographic mosaic images using fractals generated via evolutionary techniques. A photomosaic is a rendering of an image performed by placing a grid of smaller images that permit the original image to be visible when viewed from a distance. The problem of selecting the smaller images is a computationally intensive one. In this study we use an evolutionary algorithm to create fractal images on demand to generate tiles of the photomosaic. A number of images and tile resolutions are tested yielding acceptable results.

DNA edit metric codes are used as labels to track the origin of sequence data. This study is the first to treat sophisticated decoders for these error-correcting codes. Side effect machines can provide efficient decoding algorithms for such codes. Two methods for automatically producing decoding algorithms are presented. Side Effect Machines (SEMs), generalizations of finite state automata, are used in both. Single Classifier Machines (SCMs) use a single side effect machine to classify all words within a code. Locking Side Effect Machines (LSEMs) use multiple side effect machines to create a tree structured iterated classification. This study examines these techniques and provides new decoders for existing codes. Presented are ideas for best practises for the creation of these two types of new edit metric decoders. Codes of the form (n,M,d)4are used in testing due to their suitability for bioinformatics problems. A group of (12, 54-56, 7)4codes are used as an example of the process.

We provide a preliminary exploration of the use of genetic algorithms (GA) upon a substitution permutation network (SPN) cipher. The purpose of the exploration is to determine how to find weak keys. The size of the selected SPN created by Stinson gives a sample for showing the methodology and suitability of an attack using GA. We divide the types of keys into groups, each of which is analyzed to determine which groups are weaker. Simple genetic operators are examined to show the suitability of GA when applied to this problem. Results show the potential of GA to provide automated or computer assisted breaking of ciphers. The GA broke a subset of the keys using small input texts.