Pub Date : 2013-10-17DOI: 10.1109/CIG.2013.6633615
G. Greenwood
How and why cooperation develops in human populations is not known. The iterated prisoner's dilemma game provides a natural framework for studying cooperation growth in human populations. However, recent experiments with human subjects has exposed a number of serious flaws in virtually all of the game-theoretical models that have appeared in the literature. Indeed, some experiments suggest network reciprocity-thought to be essential for cooperation in human populationsmay actually play no role whatsoever. In this paper we briefly review some human experiments that were conducted in the last three years. We then present preliminary results of a new tag-mediated model designed for studying cooperation in human populations. The model exhibits many characteristics found in the human experiments including assortment, which many researchers now believe is necessary for maintaining cooperation.
{"title":"A tag-mediated game designed to study cooperation in human populations","authors":"G. Greenwood","doi":"10.1109/CIG.2013.6633615","DOIUrl":"https://doi.org/10.1109/CIG.2013.6633615","url":null,"abstract":"How and why cooperation develops in human populations is not known. The iterated prisoner's dilemma game provides a natural framework for studying cooperation growth in human populations. However, recent experiments with human subjects has exposed a number of serious flaws in virtually all of the game-theoretical models that have appeared in the literature. Indeed, some experiments suggest network reciprocity-thought to be essential for cooperation in human populationsmay actually play no role whatsoever. In this paper we briefly review some human experiments that were conducted in the last three years. We then present preliminary results of a new tag-mediated model designed for studying cooperation in human populations. The model exhibits many characteristics found in the human experiments including assortment, which many researchers now believe is necessary for maintaining cooperation.","PeriodicalId":158902,"journal":{"name":"2013 IEEE Conference on Computational Inteligence in Games (CIG)","volume":"12 2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116651961","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-10-17DOI: 10.1109/CIG.2013.6633662
K. Nguyen, Zhe Wang, R. Thawonmas
Real-Time Strategy (RTS) games typically take place in a war-like setting and are accompanied with complicated game play. They are not only difficult for human players to master, but also provide a challenging platform for AI research. In a typical RTS game, such as StarCraft, WarCraft, or Age of Empires, knowing what the opponent is doing is a great advantage and sometimes an important key to win the game. For that, good scouting is required. As subsequent work for improving the scouting agent in our StarCraft AI bot-IceBot-the winner of the mixed division in Student StarCraft AI Tournament 2012, this paper proposes a method that applies potential flows to controlling scout units in StarCraft. The proposed method outperforms an existing scouting method as well as a modified version of this existing method and is comparable to scouting by human players.
{"title":"Potential flows for controlling scout units in StarCraft","authors":"K. Nguyen, Zhe Wang, R. Thawonmas","doi":"10.1109/CIG.2013.6633662","DOIUrl":"https://doi.org/10.1109/CIG.2013.6633662","url":null,"abstract":"Real-Time Strategy (RTS) games typically take place in a war-like setting and are accompanied with complicated game play. They are not only difficult for human players to master, but also provide a challenging platform for AI research. In a typical RTS game, such as StarCraft, WarCraft, or Age of Empires, knowing what the opponent is doing is a great advantage and sometimes an important key to win the game. For that, good scouting is required. As subsequent work for improving the scouting agent in our StarCraft AI bot-IceBot-the winner of the mixed division in Student StarCraft AI Tournament 2012, this paper proposes a method that applies potential flows to controlling scout units in StarCraft. The proposed method outperforms an existing scouting method as well as a modified version of this existing method and is comparable to scouting by human players.","PeriodicalId":158902,"journal":{"name":"2013 IEEE Conference on Computational Inteligence in Games (CIG)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128898306","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-10-17DOI: 10.1109/CIG.2013.6633631
Amit Benbassat, M. Sipper
We present EvoMCTS, a genetic programming method for enhancing level of play in games. Our work focuses on the zero-sum, deterministic, perfect-information board game of Reversi. Expanding on our previous work on evolving board-state evaluation functions for alpha-beta search algorithm variants, we now evolve evaluation functions that augment the MTCS algorithm. We use strongly typed genetic programming, explicitly defined introns, and a selective directional crossover method. Our system regularly evolves players that outperform MCTS players that use the same amount of search. Our results prove scalable and EvoMCTS players whose search is increased offline still outperform MCTS counterparts. To demonstrate the generality of our method we apply EvoMCTS successfully to the game of Dodgem.
{"title":"EvoMCTS: Enhancing MCTS-based players through genetic programming","authors":"Amit Benbassat, M. Sipper","doi":"10.1109/CIG.2013.6633631","DOIUrl":"https://doi.org/10.1109/CIG.2013.6633631","url":null,"abstract":"We present EvoMCTS, a genetic programming method for enhancing level of play in games. Our work focuses on the zero-sum, deterministic, perfect-information board game of Reversi. Expanding on our previous work on evolving board-state evaluation functions for alpha-beta search algorithm variants, we now evolve evaluation functions that augment the MTCS algorithm. We use strongly typed genetic programming, explicitly defined introns, and a selective directional crossover method. Our system regularly evolves players that outperform MCTS players that use the same amount of search. Our results prove scalable and EvoMCTS players whose search is increased offline still outperform MCTS counterparts. To demonstrate the generality of our method we apply EvoMCTS successfully to the game of Dodgem.","PeriodicalId":158902,"journal":{"name":"2013 IEEE Conference on Computational Inteligence in Games (CIG)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128423909","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-10-17DOI: 10.1109/CIG.2013.6633657
Mark Grimes, M. Dror
Goofspiel is a zero-sum two player card game in which all information is known by both players. Many strategies exist that leverage random, deterministic, and learning approaches to play, however, no strategy dominates all others. It has been suggested that a hybrid strategy combining two or more of these approaches may provide better results than any of these alone. In this note, we review the strengths and weaknesses of each traditional strategy and make a cursory evaluation of a hybrid `Good' strategy.
{"title":"Observations on strategies for Goofspiel","authors":"Mark Grimes, M. Dror","doi":"10.1109/CIG.2013.6633657","DOIUrl":"https://doi.org/10.1109/CIG.2013.6633657","url":null,"abstract":"Goofspiel is a zero-sum two player card game in which all information is known by both players. Many strategies exist that leverage random, deterministic, and learning approaches to play, however, no strategy dominates all others. It has been suggested that a hybrid strategy combining two or more of these approaches may provide better results than any of these alone. In this note, we review the strengths and weaknesses of each traditional strategy and make a cursory evaluation of a hybrid `Good' strategy.","PeriodicalId":158902,"journal":{"name":"2013 IEEE Conference on Computational Inteligence in Games (CIG)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132669911","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-10-17DOI: 10.1109/CIG.2013.6633667
Ho-Chul Cho, Kyung-Joong Kim
When you prepare an entry for the StarCraft AI competitions, it is important to understand the difference between human leagues and AI bots' competitions. Simply, you can watch a lot of replays from the two leagues and compare their plays. Unfortunately, it takes much time to review them and also requires expertise for the game. Recently, it is possible to access a lot of replay files from the Internet for the two leagues. In this paper, we propose to use replay-based data mining algorithms to identify the difference between human and AI bots. It shows that the AI league has unique property compared with the human competition.
{"title":"Comparison of human and AI bots in StarCraft with replay data mining","authors":"Ho-Chul Cho, Kyung-Joong Kim","doi":"10.1109/CIG.2013.6633667","DOIUrl":"https://doi.org/10.1109/CIG.2013.6633667","url":null,"abstract":"When you prepare an entry for the StarCraft AI competitions, it is important to understand the difference between human leagues and AI bots' competitions. Simply, you can watch a lot of replays from the two leagues and compare their plays. Unfortunately, it takes much time to review them and also requires expertise for the game. Recently, it is possible to access a lot of replay files from the Internet for the two leagues. In this paper, we propose to use replay-based data mining algorithms to identify the difference between human and AI bots. It shows that the AI league has unique property compared with the human competition.","PeriodicalId":158902,"journal":{"name":"2013 IEEE Conference on Computational Inteligence in Games (CIG)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129989000","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-10-17DOI: 10.1109/CIG.2013.6633648
J. Bishop, R. Miikkulainen
Most successful examples of Reinforcement Learning (RL) report the use of carefully designed features, that is, a representation of the problem state that facilitates effective learning. The best features cannot always be known in advance, creating the need to evaluate more features than will ultimately be chosen. This paper presents Temporal Difference Feature Evaluation (TDFE), a novel approach to the problem of feature evaluation in an online RL agent. TDFE combines value function learning by temporal difference methods with an evolutionary algorithm that searches the space of feature subsets, and outputs franking over all individual features. TDFE dynamically adjusts its ranking, avoids the sample complexity multiplier of many population-based approaches, and works with arbitrary feature representations. Online learning experiments are performed in the game of Connect Four, establishing (i) that the choice of features is critical, (ii) that TDFE can evaluate and rank all the available features online, and (iii) that the ranking can be used effectively as the basis of dynamic online feature selection.
{"title":"Evolutionary Feature Evaluation for Online Reinforcement Learning","authors":"J. Bishop, R. Miikkulainen","doi":"10.1109/CIG.2013.6633648","DOIUrl":"https://doi.org/10.1109/CIG.2013.6633648","url":null,"abstract":"Most successful examples of Reinforcement Learning (RL) report the use of carefully designed features, that is, a representation of the problem state that facilitates effective learning. The best features cannot always be known in advance, creating the need to evaluate more features than will ultimately be chosen. This paper presents Temporal Difference Feature Evaluation (TDFE), a novel approach to the problem of feature evaluation in an online RL agent. TDFE combines value function learning by temporal difference methods with an evolutionary algorithm that searches the space of feature subsets, and outputs franking over all individual features. TDFE dynamically adjusts its ranking, avoids the sample complexity multiplier of many population-based approaches, and works with arbitrary feature representations. Online learning experiments are performed in the game of Connect Four, establishing (i) that the choice of features is critical, (ii) that TDFE can evaluate and rank all the available features online, and (iii) that the ranking can be used effectively as the basis of dynamic online feature selection.","PeriodicalId":158902,"journal":{"name":"2013 IEEE Conference on Computational Inteligence in Games (CIG)","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126505823","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-10-17DOI: 10.1109/CIG.2013.6633659
J. A. Brown
Presented is an overview of an evolutionary algorithm with interactive fitness evaluation as a method for development of video game weaponry, with an emphasis on games with role playing game (RPG) elements. After a short survey of current industry practises for video game weapons, an evaluation of the novel evolutionary method forms the body of this monograph. The method uses the crossover of weapon data structures of similar weapon classes. The player attempting this crossover is shown two possible weapons and is allowed to save only one. This acts as an interactive fitness/selection operator. Such a process, over the course of a game with many item drops, approximates the results of an evolutionary search. The proposed method allows for an increased engagement on the part of the player in their weapons, armour, and gear. Finally, areas for both industry applications of this technique and potential academic research topics for game balance are speculated upon.
{"title":"Evolved weapons for RPG drop systems","authors":"J. A. Brown","doi":"10.1109/CIG.2013.6633659","DOIUrl":"https://doi.org/10.1109/CIG.2013.6633659","url":null,"abstract":"Presented is an overview of an evolutionary algorithm with interactive fitness evaluation as a method for development of video game weaponry, with an emphasis on games with role playing game (RPG) elements. After a short survey of current industry practises for video game weapons, an evaluation of the novel evolutionary method forms the body of this monograph. The method uses the crossover of weapon data structures of similar weapon classes. The player attempting this crossover is shown two possible weapons and is allowed to save only one. This acts as an interactive fitness/selection operator. Such a process, over the course of a game with many item drops, approximates the results of an evolutionary search. The proposed method allows for an increased engagement on the part of the player in their weapons, armour, and gear. Finally, areas for both industry applications of this technique and potential academic research topics for game balance are speculated upon.","PeriodicalId":158902,"journal":{"name":"2013 IEEE Conference on Computational Inteligence in Games (CIG)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115495078","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-10-17DOI: 10.1109/CIG.2013.6633639
Atif M. Alhejali, S. Lucas
Ms Pac-Man is one of the most challenging test beds in game artificial intelligence (AI). Genetic programming and Monte Carlo Tree Search (MCTS) have already been successful applied to several games including Pac-Man. In this paper, we use Monte Carlo Tree Search to create a Ms Pac-Man playing agent before using genetic programming to enhance its performance by evolving a new default policy to replace the random agent used in the simulations. The new agent with the evolved default policy was able to achieve an 18% increase on its average score over the agent with random default policy.
{"title":"Using genetic programming to evolve heuristics for a Monte Carlo Tree Search Ms Pac-Man agent","authors":"Atif M. Alhejali, S. Lucas","doi":"10.1109/CIG.2013.6633639","DOIUrl":"https://doi.org/10.1109/CIG.2013.6633639","url":null,"abstract":"Ms Pac-Man is one of the most challenging test beds in game artificial intelligence (AI). Genetic programming and Monte Carlo Tree Search (MCTS) have already been successful applied to several games including Pac-Man. In this paper, we use Monte Carlo Tree Search to create a Ms Pac-Man playing agent before using genetic programming to enhance its performance by evolving a new default policy to replace the random agent used in the simulations. The new agent with the evolved default policy was able to achieve an 18% increase on its average score over the agent with random default policy.","PeriodicalId":158902,"journal":{"name":"2013 IEEE Conference on Computational Inteligence in Games (CIG)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132120207","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-10-17DOI: 10.1109/CIG.2013.6633614
Kevin Norris, I. Watson
An approximate Nash equilibrium strategy is difficult for opponents of all skill levels to exploit, but it is not able to exploit opponents. Opponent modeling strategies on the other hand provide the ability to exploit weak players, but have the disadvantage of being exploitable to strong players. We examine the effects of combining an approximate Nash equilibrium strategy with an opponent based strategy. We present a statistical exploitation module that is capable of adding opponent based exploitation to any base strategy for playing No Limit Texas Hold'em. This module is built to recognize statistical anomalies in the opponent's play and capitalize on them through the use of expert designed statistical exploitations. Expert designed statistical exploitations ensure that the addition of the module does not increase the exploitability of the base strategy. The merging of an approximate Nash equilibrium strategy with the statistical exploitation module has shown promising results in our initial experiments against a range of static opponents with varying exploitabilities. It could lead to a champion level player once the module is improved to deal with dynamic opponents.
{"title":"A statistical exploitation module for Texas Hold'em: And it's benefits when used with an approximate nash equilibrium strategy","authors":"Kevin Norris, I. Watson","doi":"10.1109/CIG.2013.6633614","DOIUrl":"https://doi.org/10.1109/CIG.2013.6633614","url":null,"abstract":"An approximate Nash equilibrium strategy is difficult for opponents of all skill levels to exploit, but it is not able to exploit opponents. Opponent modeling strategies on the other hand provide the ability to exploit weak players, but have the disadvantage of being exploitable to strong players. We examine the effects of combining an approximate Nash equilibrium strategy with an opponent based strategy. We present a statistical exploitation module that is capable of adding opponent based exploitation to any base strategy for playing No Limit Texas Hold'em. This module is built to recognize statistical anomalies in the opponent's play and capitalize on them through the use of expert designed statistical exploitations. Expert designed statistical exploitations ensure that the addition of the module does not increase the exploitability of the base strategy. The merging of an approximate Nash equilibrium strategy with the statistical exploitation module has shown promising results in our initial experiments against a range of static opponents with varying exploitabilities. It could lead to a champion level player once the module is improved to deal with dynamic opponents.","PeriodicalId":158902,"journal":{"name":"2013 IEEE Conference on Computational Inteligence in Games (CIG)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127291631","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2013-10-17DOI: 10.1109/CIG.2013.6633609
R. Sifa, C. Bauckhage
The problem of creating believable game AI poses numerous challenges for computational intelligence research. A particular challenge consists in creating human-like behaving game bots by means of applying machine learning to game-play data recorded by human players. In this paper, we propose a novel, biologically inspired approach to behavior learning for video games. Our model is based on the idea of movement primitives and we use Archetypal Analysis to determine elementary movements from data in order to represent any player action in terms of convex combinations of archetypal motions. Given these representations, we use supervised learning in order to create a system that is able to synthesize appropriate motion behavior during a game. We apply our model to teach a first person shooter game bot how to navigate in a game environment. Our results indicate that the model is able to simulate human-like behavior at lower computational costs than previous approaches.
{"title":"Archetypical motion: Supervised game behavior learning with Archetypal Analysis","authors":"R. Sifa, C. Bauckhage","doi":"10.1109/CIG.2013.6633609","DOIUrl":"https://doi.org/10.1109/CIG.2013.6633609","url":null,"abstract":"The problem of creating believable game AI poses numerous challenges for computational intelligence research. A particular challenge consists in creating human-like behaving game bots by means of applying machine learning to game-play data recorded by human players. In this paper, we propose a novel, biologically inspired approach to behavior learning for video games. Our model is based on the idea of movement primitives and we use Archetypal Analysis to determine elementary movements from data in order to represent any player action in terms of convex combinations of archetypal motions. Given these representations, we use supervised learning in order to create a system that is able to synthesize appropriate motion behavior during a game. We apply our model to teach a first person shooter game bot how to navigate in a game environment. Our results indicate that the model is able to simulate human-like behavior at lower computational costs than previous approaches.","PeriodicalId":158902,"journal":{"name":"2013 IEEE Conference on Computational Inteligence in Games (CIG)","volume":"165 12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127565371","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：