Up: Len's Home page

Machine Reasoning

Instructor: Lenhart Schubert

N.B.: Updating of this site is currently in progress

Classes: Tuesdays and Thursdays, 9:40am-10:55am, Hylan 102

Office Hours: Tues 5:00-6:00pm & Thu 5:00-6:00pm

Office: WEGMANS 3003

e-mail: my-last-name at cs dot rochester dot edu

Grad TA: Yifan Zhu (FirstLast at rochester.edu). TA's 244/444 home page
Assignment submission: uploads.cs444.2023@proton.me

UG TA: Qianqian Wei (q last-name 2 at u dot etc)

TA Office hours (Yifan Zhu): Tues 2:00-3:00pm, Fri. 4:00-5:00pm, WH 2209

Goals of the course

The ability to reason and plan based on acquired knowledge is perhaps the most distinctive mark of intelligence -- yet is the weakest faculty of current AI systems, including large language models like ChatGPT. The main goal of this course is to learn about techniques for explicitly representing factual knowledge (typically, the sort of knowledge that we can easily express in ordinary language) in a way that enables sound reasoning and effective planning towards desired goals. A central tool will be first-order logic, augmented with specialized techniques for reasoning in the face of uncertainty, and reasoning about important classes of relations such as temporal, taxonomic, and partonomic ones. Recent efforts to enable systems based on neural networks (NNs) to reason and plan, or conversely, to support symbolic reasoning and planning systems with NN technology, will be covered as well.

Practically speaking, students will reach the point where they can write programs for basic reasoning and planning tasks, and make sense of the current research literature in this area; CSC 444 students will also learn about a subtopic of their choice in greater depth, by completing a project based on reading some recent papers or books and (perhaps) writing programs that implement some ideas they have.

Prerequisites

The standard prerequisites are CSC 173 (computation & formal systems) and CSC 242 (a general course on AI). A solid math background and basic knowledge of logic and programming may suffice as well (check with instructor). Some prior exposure to functional programming (esp. Scheme, Lisp or Haskell, or basics of lambda-calculus) can be advantageous.

Credit

The grade for the course will be based on homeworks assigned every other week, plus three Lisp programming assignments, a midterm and final test, and, for 444 students, a research/essay project. See the revised course schedule for % weights. For any further grading details, see the grad TA's homepage for 244/444 (when it becomes available).

Here is a link to a pdf file providing guidelines for the graduate project (to be brought up to date by Sep. 15).

The text and supplementary texts are

The following are a couple of brief in-house introductions to Common Lisp:
         Lisp Intro (LKS),    Lisp Tutorial (GK)

Here is a lecture-by-lecture Revised Course Schedule. The assignment hand-out and turn-in dates are firm.

Latex'd notes will be provided for a majority of the topics in the class, along with more sketchy slides.
Here is a link to the Lecture Notes, to be gradually developed; and here are corresponding Slides.

Some supplementary notes and interesting links: