This Unit introduces symbolic approaches in AI. These approaches
dominated the AI field from the birth of AI until about mid-1990s, when
machine learning started to take the center stage. Although they are
brittle and do not scale, symbolic approaches also enjoy advantages that
machine learning (esp. deep learning) methods do not possess, such as
explainability and transparency, provable verification, efficient
algorithms (such as A* search and dynamic programming), and linguistic
relevance.
Module Learning Outcomes
After successful completion of this module, you should be able to do
the following (in addition to answering the questions listed below):
Identify scenarios that are particularly suitable for symbolic AI
(CLO 1)
What problem domains are particularly friendly to symbolic AI?
Formulate real-world puzzles as search problems (CLO 1)
What is a search state?
What is a search graph?
What algorithms are used to search for a path?
Why heuristic search (like A*) is faster than uninformed
search?
Formulate game AI as adversarial search (CLO 1)
What is a minimax search tree?
What is an evaluation function?
Explain context-free grammars and parsing (CLO 1)
What are nonterminals, preterminals, and terminals in a context-free
grammar?
What are rewriting rules? Give an example.
Draw two parse trees for “I saw a man with a telescope”.
How does CKY algorithm work, in a high-level?
How are the probabilities learned from data? What kind of data?