Font Size: a A A

The Encoding and Use of Abstract Visual Representations: Statistical Learning, Rule Learning, and Object Recognition

Posted on:2013-06-24Degree:Ph.DType:Dissertation
University:Brandeis UniversityCandidate:MacKenzie, Kimberly JFull Text:PDF
GTID:1458390008975080Subject:Neurosciences
Abstract/Summary:
In this dissertation I examine how visual representations are constructed and consolidated for later use by adults and infants. A general introduction to the many facets of visual learning is provided by Chapter 1, as well as an explanation for why observational learning is ideal for forming internal representations. In Chapter 2, I demonstrate the role two forms of observational visual learning, statistical and rule, play in the creation of visual representations. The two forms of learning are found to interact, as the learning of a statistical structure can assist in the learning of a rule structure. These results indicate that statistical and rule learning may not be separate learning mechanisms, as has been previously suggested.;In Chapter 3, I explore how the memories formed by statistical and rule learning are consolidated into long term representations, specifically whether sleep assists in the consolidation process. I show that while statistical structures are retained in long term memory, sleep does not offer a benefit to performance, unlike many skill learning tasks.;In Chapter 4, I examine the scope of observational learning, by determining the extent to which an infant can extract embedded statistical structures in complex scenes. Infant performance was similar to that of adults, in that embedded structures are ignored in favor of larger structures. An infant, however, is less capable of abstraction than an adult.;In Chapter 5, I investigate how visual representations are used in processes such as object recognition and categorization, by examining how visual information is integrated, based on performance. Object categorization results show that receiving a high amount of information, such as from fixation, results in better performance in an object categorization task, than when a lower amount of information, such as from peripheral vision is received first. This result indicates information integration from high to low results in greater access to a stored representation.;Overall, these studies give new evidence as to how visual representations are created by adults and infants, and how they are stored and employed. They also support the role of observational learning as the manner in which abstract representations are formed.
Keywords/Search Tags:Representations, Statistical, Rule learning, Observational learning, Object
Related items