16.01.2015 Views

CS2013-final-report

CS2013-final-report

CS2013-final-report

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

CS 662; Artificial Intelligence Programming, University of San<br />

Francisco<br />

Christopher Brooks<br />

cbrooks@usfca.edu<br />

https://sierra.cs.usfca.edu<br />

Knowledge Areas that contain topics and learning outcomes covered in the course<br />

Knowledge Area<br />

Total Hours of Coverage<br />

Intelligent Systems (IS) 60<br />

Where does the course fit in your curriculum<br />

The course is taken as a senior-level elective, and as a required course for first-year Master’s students.<br />

What is covered in the course<br />

An overview of AI, including search, knowledge representation, probabilistic reasoning and decision making<br />

under uncertainty, machine learning, and topics from NLP, information retrieval, knowledge engineering and<br />

multi-agent systems.<br />

What is the format of the course<br />

Face-to-face. 4 hours lecture per week (either 3x65 or 2x105).<br />

How are students assessed<br />

Two midterms and a <strong>final</strong>. Also, heavy emphasis on programming and implementation of techniques. Students<br />

complete 7-9 assignments (see website for examples). Expectation is 3 hours outside of class for every hour of<br />

lecture.<br />

Course textbooks and materials<br />

Russell and Norvig’s Artificial Intelligence: A Modern Approach is the primary text. I prepare lots of summary<br />

material (see website) and provide students with harness code for their assignments in Python. I also make use of<br />

pre-existing packages and tools such as NLTK, Protégé and WordNet when possible.<br />

Why do you teach the course this way<br />

My goals are:<br />

• Illustrate the ways in which AI techniques can be used to solve real-world problems. I pick a specific<br />

domain (such as Wikipedia or a map of San Francisco) and have the students apply a variety of<br />

techniques to problems in this domain. For example, as the assumptions change, the same SF map<br />

problem can be used for search, constraints, MDPs, planning, or learning.<br />

• Provide students with experience implementing these algorithms. Almost all of our students go into<br />

industry, and need skill in building systems.<br />

• Illustrate connections between techniques. For example, I discuss decision trees and rule learning in<br />

conjunction with forward and backward chaining to show how, once you’ve decided on a representation,<br />

you can either construct rules using human expertise, or else learn them (or both).<br />

- 304 -

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!