AS 050.480 - Learning Theory

Institution:

Johns Hopkins University

Subject:

Description:

Prerequisites: With instructor permission, this course is open to upperclass undergraduates concentrating in computation. Recently, statistical learning has played a leading role in informing the empiricist/nativist and connectionist/ symbolic debates. But just what is “statistical learning” and what’s new about it? This course presents theories of statistical learning, such as Bayesian models, causal networks, information-theoretic models (e.g., Minimum Description Length and Maximum Entropy formalisms). These methods have caused revolutions in machine vision and natural language processing. During the course, these methods will be compared with other numerical learning methods such as connectionist networks, and with non-numerical learning theories such as Gold’s classic learnability theory and its probailistic extension to PAC (probably approximately correct) learning theory. This recent work has fundamental implications for the ancient problem of induction. Prerequisites: With instructor permission, this course is open to upperclass undergraduates concentrating in computation.

Credits:

3.00

Credit Hours:

Prerequisites:

Corequisites:

Exclusions:

Level:

Instructional Type:

Lecture

Notes:

Additional Information:

Historical Version(s):

Institution Website:

www.jhu.edu

Phone Number:

(410) 516-8000

Regional Accreditation:

Middle States Association of Colleges and Schools

Calendar System:

Semester

The Course Profile information is provided and updated by third parties including the respective institutions. While the institutions are able to update their information at any time, the information is not independently validated, and no party associated with this website can accept responsibility for its accuracy.