About
46
Publications
7,178
Reads
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
538
Citations
Citations since 2017
Introduction
Publications
Publications (46)
Generative Adversarial Networks (GANs) are a machine learning approach capable of generating novel example outputs across a space of provided training examples. Procedural Content Generation (PCG) of levels for video games could benefit from such models, especially for games where there is a pre-existing corpus of levels to emulate. This paper trai...
Game balancing is an important part of the (computer) game design process, in which designers adapt a game prototype so that the resulting gameplay is as entertaining as possible. In industry, the evaluation of a game is often based on costly playtests with human players. It suggests itself to automate this process using surrogate models for the pr...
Uncertainty propagation is a technique to incorporate individuals with uncertain fitness estimates in evolutionary algorithms. The Surrogate-Assisted Partial Order-Based Evolutionary Optimisation Algorithm (SAPEO) uses uncertainty propagation of fitness predictions from a Kriging model to reduce the number of function evaluations. The fitness predi...
The term Procedural Content Generation (PCG) refers to the (semi-)automatic generation of game content by algorithmic means, and its methods are becoming increasingly popular in game-oriented research and industry. A special class of these methods, which is commonly known as search-based PCG, treats the given task as an optimisation problem. Such p...
Generative Adversarial Networks (GANs) are a powerful indirect genotype-to-phenotype mapping for evolutionary search. Much previous work applying GANs to level generation focuses on fixed-size segments combined into a whole level, but individual segments may not fit together cohesively. In contrast, segments in human designed levels are often repea...
Generative Adversarial Networks (GANs) are a powerful indirect genotype-to-phenotype mapping for evolutionary search, but they have limitations. In particular, GAN output does not scale to arbitrary dimensions, and there is no obvious way to combine GAN outputs into a cohesive whole, which would be useful in many areas, such as video game level gen...
Optimisation algorithms are commonly compared on benchmarks to get insight into performance differences. However, it is not clear how closely benchmarks match the properties of real-world problems because these properties are largely unknown. This work investigates the properties of real-world problems through a questionnaire to enable the design o...
Benchmarks are a useful tool for empirical performance comparisons. However, one of the main shortcomings of existing benchmarks is that it remains largely unclear how they relate to real-world problems. What does an algorithm’s performance on a benchmark say about its potential on a specific real-world problem? This work aims to identify propertie...
Benchmarks are a useful tool for empirical performance comparisons. However, one of the main shortcomings of existing benchmarks is that it remains largely unclear how they relate to real-world problems. What does an algorithm’s performance on a benchmark say about its potential on a specific real-world problem? This work aims to identify propertie...
This survey compiles ideas and recommendations from more than a dozen researchers with different backgrounds and from different institutes around the world. Promoting best practice in benchmarking is its main goal. The article discusses eight essential topics in benchmarking: clearly stated goals, well-specified problems, suitable algorithms, adequ...
While games have been used extensively as milestones to evaluate game-playing AI, there exists no standardised framework for reporting the obtained observations. As a result, it remains difficult to draw general conclusions about the strengths and weaknesses of different game-playing AI algorithms. In this paper, we propose reporting guidelines for...
Recent procedural content generation via machine learning (PCGML) methods allow learning from existing content to produce similar content automatically. While these approaches are able to generate content for different games (e.g. Super Mario Bros., DOOM, Zelda, and Kid Icarus), it is an open questions how well these approaches can capture large-sc...
Benchmarks are a useful tool for empirical performance comparisons. However, one of the main shortcomings of existing benchmarks is that it remains largely unclear how they relate to real-world problems. What does an algorithm's performance on a benchmark say about its potential on a specific real-world problem? This work aims to identify propertie...
Generative Adversarial Networks (GANs) are proving to be a powerful indirect genotype-to-phenotype mapping for evolutionary search, but they have limitations. In particular, GAN output does not scale to arbitrary dimensions, and there is no obvious way of combining multiple GAN outputs into a cohesive whole, which would be useful in many areas, suc...
Generative Adversarial Networks (GANs) are an emerging form of indirect encoding. The GAN is trained to induce a latent space on training data, and a real-valued evolutionary algorithm can search that latent space. Such Latent Variable Evolution (LVE) has recently been applied to game levels. However, it is hard for objective scores to capture leve...
This paper examines learning approaches for forward models based on local cell transition functions. We provide a formal definition of local forward models for which we propose two basic learning approaches. Our analysis is based on the game Sokoban, where a wrong action can lead to an unsolvable game state. Therefore, an accurate prediction of an...
Real-world problems are often affected by uncertainties of different types and from multiple sources. Algorithms created for expensive optimisation, such as model-based optimisers, introduce additional errors. We argue that these uncertainties should be accounted for during the optimisation process. We thus introduce a benchmark as well as a new su...
This paper provides a detailed investigation of using the Kullback-Leibler (KL) Divergence as a way to compare and analyse game-levels, and hence to use the measure as the objective function of an evolutionary algorithm to evolve new levels. We describe the benefits of its asymmetry for level analysis and demonstrate how (not surprisingly) the qual...
Despite a large interest in real-world problems from the research field of evolutionary optimisation, established benchmarks in the field are mostly artificial. We propose to use game optimisation problems in order to form a benchmark and implement function suites designed to work with the established COCO benchmarking framework. Game optimisation...
Hyperparameter tuning is an important mixed-integer optimisation problem, especially in the context of real-world applications such as games. In this paper, we propose a function suite around hyperparameter optimisation of game AI based on the card game Splendor and using the Rinascimento framework. We propose several different functions and demons...
In this position paper, we discuss the need for systematic benchmarking of surrogate-assisted evolutionary algorithms and give an overview of existing suitable function suites. Based on the findings, we hope to encourage more comparative studies in this field supported by benchmarks and outline how a concerted effort of the community could create b...
Realtime strategy games (and especially StarCraft II) are currently becoming the ‘next big thing’ in Game AI, as building human competitive bots for complex games is still not possible. However, the abundance of existing game data makes StarCraft II an ideal testbed for machine learning. We attempt to use this for establishing winner predictors tha...
This paper provides a detailed investigation of using the Kullback-Leibler (KL) Divergence as a way to compare and analyse game-levels, and hence to use the measure as the objective function of an evolutionary algorithm to evolve new levels. We describe the benefits of its asymmetry for level analysis and demonstrate how (not surprisingly) the qual...
This paper investigates the effect of learning a forward model on the performance of a statistical forward planning agent. We transform Conway's Game of Life simulation into a single-player game where the objective can be either to preserve as much life as possible or to extinguish all life as quickly as possible. In order to learn the forward mode...
This paper introduces a simple and fast variant of Planet Wars as a test-bed for statistical planning based Game AI agents, and for noisy hyper-parameter optimisation. Planet Wars is a real-time strategy game with simple rules but complex game-play. The variant introduced in this paper is designed for speed to enable efficient experimentation, and...
Since its inauguration in 1966, the ACM A. M. Turing Award has recognized major contributions of lasting importance in computing. Through the years, it has become the most prestigious technical award in the field, often referred to as the "Nobel Prize of computing." During 2017, ACM celebrated 50 years of the Turing Award and the visionaries who ha...
Surrogate-Assisted Evolutionary Algorithms (SAEAs) have been proven
to be very effective in solving (synthetic and real-world) computationally expensive optimization problems with a limited number of function evaluations. The two main components of SAEAs are: the surrogate model and the evolutionary optimizer, both of which use parameters to contro...
Generative Adversarial Networks (GANs) are a machine learning approach capable of generating novel example outputs across a space of provided training examples. Procedural Content Generation (PCG) of levels for video games could benefit from such models, especially for games where there is a pre-existing corpus of levels to emulate. This paper trai...
Although navigation is a frequent problem in games, we find that it is rarely targeted for complex applications in scientific research. In this paper, we present multiple solutions to four major challenges for advanced navigation problems. We demonstrate, that by combining these methods, we are able to solve the navigation problem for flying agents...
ACM-W provides support for women undergraduate and graduate students in Computer Science and related programs to attend research conferences. This exposure to the CS research world can encourage a student to continue on to the next level (Undergraduate to Graduate, Masters to Ph.D, Ph.D. to an industry or academic position). The student does not ha...
In this paper, we propose a novel approach (SAPEO) to support the survival selection process in evolutionary multi-objective algorithms with surrogate models. The approach dynamically chooses individuals to evaluate exactly based on the model uncertainty and the distinctness of the population. We introduce multiple SAPEO variants that differ in ter...
Game balancing is a recurring problem that currently requires a lot of manual work, usually following a game designer’s intuition or rules-of-thumb. To what extent can or should the balancing process be automated? We establish a process model that integrates both manual and automated balancing approaches. Artificial agents are employed to automatic...
In the General Video Game Playing competitions of the last years, Monte-Carlo tree search as well as Evolutionary Algorithm based controllers have been successful. However, both approaches have certain weaknesses, suggesting that certain hybrids could outperform both. We envision and experimentally compare several types of hybrids of two basic appr...
Crowdfunding, a strategy for obtaining financial support for a project, an event, or a start-up company, has become both popular and successful in recent years. However, research on the subject is still scarce. Based on data from Kickstarter (US-based) and Startnext (Germany-based) projects, insights into the current crowdfunding situation at the a...
Projects
Projects (3)
Our goal is to clarify the characteristics of real-world optimization problems and generate a new benchmark test suite based on the questionnaire (https://tinyurl.com/opt-survey) results. We believe such a test suite helps efficiently developing new optimization methods for real-world optimization problems. Results will be made publicly available at https://sites.google.com/view/macoda-rwp
I work on the automatic configuration of systems with human agents using the example of game balancing through surrogate-assisted multi-objective evolutionary optimisation. The idea is to (1) define measurable balancing goals (player preferences, designer intention) of a game, (2) set up a simulation of the game using (believable and generalisable) AIs and then (3) optimise a set of parameters within the rules of the game. The fitness is evaluated based on the simulation, a surrogate model or playtests with human players. The questions I'm most interested in researching are the effects of the simulation accuracy and the surrogate model on an optimisation algorithm that uses them and the existence of generalisable patterns of the fitness landscape. Eventually, the goal is to see if the obtained results can be transferred to real-world problems, where simulations are necessary as well, but might be much more costly and more difficult to verify.
