In this studythe author investigates information processing in deep autoencoder models. It is demonstrated that unsupervised training of autoencoders of certain class can result in emergence of compact and structured internal representations of the input data space that can be correlated with higher level categories. The authors propose and demonstrate the practical possibility to detect and measure this emergent information structure by applying unsupervised density clustering in the activation space of the focal hidden layer of the model. Based on the findings of the studya new approach to training neural network models is proposed that is based on the emergent in unsupervised training information landscape, that is iterative, driven by the environment, requires minimal supervision and with interesting similarities to learning of biologic systems. In conclusion, a discussion of theoretical foundations of spontaneous categorization in self-learning systemsis provided.


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    SPONTANEOUS CATEGORIZATION AND SELF-LEARNING WITH DEEP AUTOENCODER MODELS


    Beteiligte:
    Serge Dolgikh (Autor:in)


    Erscheinungsdatum :

    2019




    Medientyp :

    Aufsatz (Zeitschrift)


    Format :

    Elektronische Ressource


    Sprache :

    Unbekannt




    Learning Compositional Categorization Models

    Ommer, B. / Buhmann, J. M. | British Library Conference Proceedings | 2006


    Learning Compositional Categorization Models

    Ommer, Björn / Buhmann, Joachim M. | Springer Verlag | 2006

    Freier Zugriff

    Movie Recommendation System Using Deep Autoencoder

    Srikanth, Purushothaman / Ushitaasree, E. / Bhargav Bhattaram, Sai Mani et al. | IEEE | 2021


    Reinforcement learning and automatic categorization

    Porta, Josep M. / Celaya, Enric | BASE | 1999

    Freier Zugriff

    VISUALIZATION OF DRIVING BEHAVIOR USING DEEP SPARSE AUTOENCODER

    Liu, H. / Taniguchi, T. / Takano, T. et al. | British Library Conference Proceedings | 2014