Filters








1,184 Hits in 7.8 sec

Interactive Evolution and Exploration Within Latent Level-Design Space of Generative Adversarial Networks [article]

Jacob Schrum, Jake Gutierrez, Vanessa Volz, Jialin Liu, Simon Lucas,, Sebastian Risi
2020 arXiv   pre-print
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.  ...  ACKNOWLEDGMENTS The authors would like to thank the Schloss Dagstuhl team and the organisers of Dagstuhl Seminars 17471 and 19511 for hosting productive seminars.  ... 
arXiv:2004.00151v1 fatcat:52wewnv2ove7lbxy7444mgvkkq

Evolving Mario Levels in the Latent Space of a Deep Convolutional Generative Adversarial Network [article]

Vanessa Volz, Jacob Schrum, Jialin Liu, Simon M. Lucas, Adam Smith, Sebastian Risi
2018 arXiv   pre-print
Generative Adversarial Networks (GANs) are a machine learning approach capable of generating novel example outputs across a space of provided training examples.  ...  Specifically, various fitness functions are used to discover levels within the latent space of the GAN that maximize desired properties.  ...  Acknowledgements The authors would like to thank the Schloss Dagstuhl team and the organisers of the Dagstuhl Seminar 17471 for a creative and productive seminar.  ... 
arXiv:1805.00728v1 fatcat:xwabyb2drjcnjjbhtpohz3oniy

Evolving mario levels in the latent space of a deep convolutional generative adversarial network

Vanessa Volz, Jacob Schrum, Jialin Liu, Simon M. Lucas, Adam Smith, Sebastian Risi
2018 Proceedings of the Genetic and Evolutionary Computation Conference on - GECCO '18  
Generative Adversarial Networks (GANs) are a machine learning approach capable of generating novel example outputs across a space of provided training examples.  ...  Specifically, various fitness functions are used to discover levels within the latent space of the GAN that maximize desired properties.  ...  Acknowledgements The authors would like to thank the Schloss Dagstuhl team and the organisers of the Dagstuhl Seminar 17471 for a creative and productive seminar.  ... 
doi:10.1145/3205455.3205517 dblp:conf/gecco/VolzSLLSR18 fatcat:eb5oi3dxxbhe7epy7l3laamfci

Bootstrapping Conditional GANs for Video Game Level Generation [article]

Ruben Rodriguez Torrado, Ahmed Khalifa, Michael Cerny Green, Niels Justesen, Sebastian Risi, Julian Togelius
2019 arXiv   pre-print
Ad-ditionally, to reduce the number of levels necessary to trainthe GAN, we propose a bootstrapping mechanism in whichplayable generated levels are added to the training set.  ...  Generative Adversarial Networks (GANs) have shown im-pressive results for image generation.  ...  Michael Cerny Green acknowledges the financial support of the GAANN program. All authors acknowledge Per Josefsen and Nicola Zaltron, who were responsible for the 45 human-designed levels.  ... 
arXiv:1910.01603v1 fatcat:6uazxl2n3ncv5imokuflvqf73e

Deep Learning for Procedural Content Generation

Jialin Liu, Sam Sndograss, Ahmed Khalifa, Sebastian Risi, Georgios N. Yannakakis, Julian Togelius
2020 Zenodo  
Existing procedural content generation methods, such as search-based, solver-based, rule-based and grammar-based methods have been applied to various content types such as levels, maps, character models  ...  More recently, deep learning has powered a remarkable range of inventions in content production, which are applicable to games.  ...  to search for content in the learned space of a GAN/VAE.  ... 
doi:10.5281/zenodo.4415242 fatcat:6q4swrsefvhhde2v6mepsoagg4

Illuminating the Space of Beatable Lode Runner Levels Produced By Various Generative Adversarial Networks [article]

Kirby Steckel, Jacob Schrum
2021 arXiv   pre-print
Generative Adversarial Networks (GANs) are capable of generating convincing imitations of elements from a training set, but the distribution of elements in the training set affects to difficulty of properly  ...  Interestingly, a GAN trained on only 20 levels generated the largest set of diverse beatable levels while a GAN trained on 150 levels generated the smallest set of diverse beatable levels, thus challenging  ...  ACKNOWLEDGMENTS This research was made possible by the donation-funded Summer Collaborative Opportunities and Experiences (SCOPE) program for undergraduate research at Southwestern University.  ... 
arXiv:2101.07868v1 fatcat:qt4rby7c3vad5m4x47ysytluua

Using Multiple Generative Adversarial Networks to Build Better-Connected Levels for Mega Man [article]

Benjamin Capps, Jacob Schrum
2021 arXiv   pre-print
Generative Adversarial Networks (GANs) can generate levels for a variety of games. This paper focuses on combining GAN-generated segments in a snaking pattern to create levels for Mega Man.  ...  Flow was further improved by evolving the latent vectors for the segments being joined in the level to maximize the length of the level's solution path.  ...  INTRODUCTION Generative Adversarial Networks (GANs [8] ) are capable of reproducing certain aspects of a given training set.  ... 
arXiv:2102.00337v2 fatcat:4hqhzjwi3fhhnfu26wjjvol42a

Generative Adversarial Network Rooms in Generative Graph Grammar Dungeons for The Legend of Zelda [article]

Jake Gutierrez, Jacob Schrum
2020 arXiv   pre-print
Generative Adversarial Networks (GANs) have demonstrated their ability to learn patterns in data and produce new exemplars similar to, but different from, their training set in several domains, including  ...  This approach is validated by a user study, showing that GAN dungeons are as enjoyable to play as a level from the original game, and levels generated with a graph grammar alone.  ...  ACKNOWLEDGMENTS This research is supported in part by the Summer Collaborative Opportunities and Experiences (SCOPE) program, funded by various donors to Southwestern University.  ... 
arXiv:2001.05065v2 fatcat:hj7ejlalqrbvvmxmlzbhps6v3m

World Models [article]

David Ha, Jürgen Schmidhuber
2018 arXiv   pre-print
We explore building generative neural network models of popular reinforcement learning environments.  ...  Our world model can be trained quickly in an unsupervised manner to learn a compressed spatial and temporal representation of the environment.  ...  , and for offering their valuable perspectives and insights from their areas of expertise.  ... 
arXiv:1803.10122v3 fatcat:qatufm66k5d6ncg7pnl2otbw2e

World Models

David Ha, Jürgen Schmidhuber
2018 Zenodo  
We explore building generative neural network models of popular reinforcement learning environments.  ...  Our world model can be trained quickly in an unsupervised manner to learn a compressed spatial and temporal representation of the environment.  ...  valuable perspectives and insights from their areas of expertise to make this article better.  ... 
doi:10.5281/zenodo.1207048 fatcat:a4txnvzg3bdgzcsnnskqcgcn34

Towards Game Design via Creative Machine Learning (GDCML) [article]

Anurag Sarkar, Seth Cooper
2020 arXiv   pre-print
However, similar creative ML techniques have not been as widely adopted in the domain of game design despite the emergence of ML-based methods for generating game content.  ...  Such creative ML approaches have seen wide use in the domains of visual art and music for applications such as image and music generation and style transfer.  ...  input segment or a label, i.e. the model generates desired levels without having to use latent search.  ... 
arXiv:2008.13548v1 fatcat:54eafqo6sng2jk7jqpj3ocitvi

Increasing Generality in Machine Learning through Procedural Content Generation [article]

Sebastian Risi, Julian Togelius
2020 arXiv   pre-print
Procedural Content Generation (PCG) refers to the practice, in videogames and other games, of generating content such as levels, quests, or characters algorithmically.  ...  Motivated by the need to make games replayable, as well as to reduce authoring burden, limit storage space requirements, and enable particular aesthetics, a large number of PCG methods have been devised  ...  Both authors (SR and JT) contributed equally to the conceptualization and writing of the paper.  ... 
arXiv:1911.13071v2 fatcat:ehbj2rrw5vbyvc4cg6zw3ctln4

Discovering representations for black-box optimization

Adam Gaier, Alexander Asteroth, Jean-Baptiste Mouret
2020 Proceedings of the 2020 Genetic and Evolutionary Computation Conference  
Our main insight is that this process can be automated by generating a dataset of high-performing solutions with a quality diversity algorithm (here, MAP-Elites), then learning a representation with a  ...  Our second insight is that this representation can be used to scale quality diversity optimization to higher dimensions -but only if we carefully mix solutions generated with the learned representation  ...  the German Federal Ministry of Education and Research (BMBF) under the Forschung an Fachhochschulen mit Unternehmen programme (grant agreement number 03FH012PX5, project "Aeromat").  ... 
doi:10.1145/3377930.3390221 dblp:conf/gecco/GaierAM20 fatcat:xgx7ibxeqbbctorqorudbgp7qa

Automating Representation Discovery with MAP-Elites [article]

Adam Gaier, Alexander Asteroth, Jean-Baptiste Mouret
2020 arXiv   pre-print
We evaluate the DDE approach by evolving solutions for inverse kinematics of a planar arm (200 joint angles) and for gaits of a 6-legged robot in action space (a sequence of 60 positions for each of the  ...  Our approach learns this data-driven encoding during optimization by balancing between exploiting the DDE to generalize the knowledge contained in the current archive of elites and exploring new representations  ...  For instance, given a dataset of face images, a variational autoencoder (VAE) [31] or a Generative Adversarial Network (GAN) [25] can learn a low-dimensional latent space, that is, a representation  ... 
arXiv:2003.04389v1 fatcat:imu4vz57cnft3kj4xlqkvfhz7a

Graph Based Wave Function Collapse Algorithm for Procedural Content Generation in Games

Hwanhee KIM, Teasung HAHN, Sookyun KIM, Shinjin KANG
2020 IEICE transactions on information and systems  
The goal of this system is to enable a game designer to procedurally create key content elements in the game level through simple association rule input.  ...  With our system, if the user inputs the minimum association rule, it is possible to effectively perform procedural content generation in the three-dimensional world.  ...  They utilized a Deep Convolutional Generative Adversarial Network (DCGAN) for learning about and generating sprites [16] .  ... 
doi:10.1587/transinf.2019edp7295 fatcat:vj74imbxrbechd34ch3daidf34
« Previous Showing results 1 — 15 out of 1,184 results