Rotation, Translation, and Cropping for Zero-Shot Generalization

Chang Ye, Ahmed Khalifa, Philip Bontrager, Julian Togelius

    Research output: Chapter in Book/Report/Conference proceedingConference contribution


    Deep Reinforcement Learning (DRL) has shown impressive performance on domains with visual inputs, in particular various games. However, the agent is usually trained on a fixed environment, e.g. a fixed number of levels. A growing mass of evidence suggests that these trained models fail to generalize to even slight variations of the environments they were trained on. This paper advances the hypothesis that the lack of generalization is partly due to the input representation, and explores how rotation, cropping and translation could increase generality. We show that a cropped, translated and rotated observation can get better generalization on unseen levels of two-dimensional arcade games from the GVGAI framework. The generality of the agents is evaluated on both human-designed and procedurally generated levels.

    Original languageEnglish (US)
    Title of host publicationIEEE Conference on Games, CoG 2020
    PublisherIEEE Computer Society
    Number of pages8
    ISBN (Electronic)9781728145334
    StatePublished - Aug 2020
    Event2020 IEEE Conference on Games, CoG 2020 - Virtual, Osaka, Japan
    Duration: Aug 24 2020Aug 27 2020

    Publication series

    NameIEEE Conference on Computatonal Intelligence and Games, CIG
    ISSN (Print)2325-4270
    ISSN (Electronic)2325-4289


    Conference2020 IEEE Conference on Games, CoG 2020
    CityVirtual, Osaka


    • A2C
    • generalization
    • gvgai
    • reinforcement learning
    • representation
    • zero-shot generalization

    ASJC Scopus subject areas

    • Artificial Intelligence
    • Computer Graphics and Computer-Aided Design
    • Computer Vision and Pattern Recognition
    • Human-Computer Interaction
    • Software


    Dive into the research topics of 'Rotation, Translation, and Cropping for Zero-Shot Generalization'. Together they form a unique fingerprint.

    Cite this