Learning via prediction : mapping continuous stimuli to discrete symbols
Abstract/Contents
- Abstract
- How do we use and represent words? How do we learn to break up the dimensions of continuous variation of the world into discrete categories? In this thesis, I explore how recasting this problem in terms of simple prediction provides insight into the computational nature of word learning. Through a series of computational simulations and human experiments that manipulate the structure of information in time, I find that when using features of the world to predict words, the representations learned are likely to be more useful for discriminating the appropriate response. However, these representations are also likely to be distorted, favoring diagnostic information, and thus sacrificing general utility across contexts. At the same time, trying to use words to predict the relevant features of the world will result in less distorted representations, but will not enhance discrimination. Demonstrating these effects while controlling for various potentially confounding variables strengthens the case that prediction is central to word learning.
Description
Type of resource | text |
---|---|
Form | electronic; electronic resource; remote |
Extent | 1 online resource. |
Publication date | 2010 |
Issuance | monographic |
Language | English |
Creators/Contributors
Associated with | November, Adam Daniel | |
---|---|---|
Associated with | Stanford University, Department of Psychology. | |
Primary advisor | Ramscar, Michael | |
Primary advisor | Thomas, Ewart A. C | |
Thesis advisor | Ramscar, Michael | |
Thesis advisor | Thomas, Ewart A. C | |
Thesis advisor | McClelland, James L | |
Advisor | McClelland, James L |
Subjects
Genre | Theses |
---|
Bibliographic information
Statement of responsibility | Adam Daniel November. |
---|---|
Note | Submitted to the Department of Psychology. |
Thesis | Thesis (Ph.D.)--Stanford University, 2010. |
Location | electronic resource |
Access conditions
- Copyright
- © 2010 by Adam Daniel November
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...