Learning the systematic transformation of holographic reduced representations |
| |
| Institution: | 1. Department of Neurology, Johns Hopkins University School of Medicine, United States;2. Department of Psychology, University of South Carolina, United States;3. Department of Communication Disorders, University of South Carolina, United States;4. Department of Neurology, Medical University of South Carolina, United States;5. Department of Speech & Hearing Science, Arizona State University, United States;6. University of California San Diego, United States;7. San Diego State University, United States;8. Departments of Cognitive Sciences & Language Science, University of California, Irvine, United States;1. Center for Interdisciplinary Brain Research, University of Jyväskylä, Finland;2. Department of Psychology, University of Jyväskylä, Finland;3. Center for Molecular and Behavioral Neuroscience, Rutgers University-Newark, NJ, USA;1. L-Universita‘ ta-Malta, Malta;2. LMU Munich, Germany;1. Department of Computer Science Electrical and Space Engineering, Luleå University of Technology, 971 87 Luleå, Sweden;2. Independent Researcher, Melbourne, VIC, Australia;3. Clayton School of Information Technology, Monash University, Clayton, VIC, Australia;4. Media and Communication School, Royal Melbourne Institute of Technology, Melbourne, VIC, Australia |
| |
| Abstract: | Holographic Reduced Representation is a representational scheme which allows for the representation of variable-sized structures in a distributed manner. It has been shown that these compositional structures can be transformed holistically. However, in order to do so, the transformation vector was constructed by hand. In this paper we present two methods of learning the holistic transformation of Holographic Reduced Representations from examples. We show that the acquired knowledge can be generalised to structures containing unseen elements and to structures more complex than the training examples. These generalisations require a degree of systematicity which to our knowledge has not yet been achieved by other comparable methods. |
| |
| Keywords: | |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
