dias Diaz
Computer Science Adventures

Knowledge-Based AI Review (CS7637)

Posted by Alejandro Diaz on May 08, 2021

Takeaways

  • The course covers classic AI algorithms like generate and test and case-based learning
  • A fantastic introduction to the program, the coursework is lighter on programming than others. There is a lot of writing, with weekly deliverables

What is Knowledge-based AI (KBAI)?

Knowledge-based AI (KBAI) is a course offered by GATECH’s OMSCS. The course deals with applying human-like cognitive techniques to develop artificial intelligence. The class provides two goals: Understand the basic architectures, representations, and techniques for building knowledge-based AI agents issues and methods of knowledge-based AI

KBAI is a fantastic introduction to the program. The class uses Python exclusively. You can succeed in the course with little to no background in computer science algorithms and mathematics (You will have to work harder!). The class shines when connecting AI techniques to human cognition. The lectures introduce theories from KBAI literature, postulating on how humans learn, what creativity is, and what humor is.

In a way, the class is subtly both a computer science course and a course on human psychology.

What did I learn?

  • Common strategies in KBAI
    • Utilizing semantic networks for problem solving
    • Methods like generate and test, means-ends analysis, Incremental concept learning, and case-based reasoning
    • Data structures like frames, discrimination trees, and indexing
  • Learning about learning
    • The course stresses how to set up an AI agent to learn and how that relates to human psychology
  • AI image perception
    • The course project is solving RPM tests (image) using an AI agent. You are given images to work with so will have to learn to use an image processing library like Pillow or OpenCV
  • Linear algebra application
    • Linear algebra was not required, but it helped when thinking about how to process images as images are represented as matrices using Python’s NumPy

What did we do?

  • Mini projects
    • Starting simple doing the sheep and wolves problem
    • Later more difficult, processing language to create meaning by giving your agent an ontology of the world
  • Explored ethics and laws concerning AI
  • Raven’s Progressive matrices

Why take this course?

  • Good introduction to the program
  • The professor is a fantastic lecturer
  • Good, hands-on review into algorithms, especially tree traversals and search algorithms
  • You get to design your own AI program!

-Well, till next time space cowboy

Alex