Topics Covered since 1st midterm… CSE 391 Fall 2008
Probability: Topics to Know Chain rule Bayes’ rule Axioms Why conditional independence is important and useful CIS 391 - Intro to AI
Probability: Skills you should have Convert a problem to a solvable form using Bayes’ Rule Given a joint distribution, answer questions about it Given a situation, build a Bayes network for it with observed, hidden and evidence variables, explaining why you make the independence assumptions you do, and show the CPTs Determine whether various things are independent or conditionally independent of each other Apply the Naïve Bayes algorithm to a very small example Map a Naïve Bayes problem into an appropriate Bayes network Do a step of the Viterbi algorithm on a very small example Explain the Forward Algorithm and the Backward Algorithm CIS 391 - Intro to AI
Perceptrons and SVMs Topics to Know: Perceptrons, (averaged perceptrons, voted perceptrons, MIRA) : what they are, and why you might want to use each of them Understand what the limits of perceptrons are Understanding “margin” How linear classifiers can be used to create non-linear classification boundaries Skills you should have: Perform a perceptron update step CIS 391 - Intro to AI
Logic: Topics to Know Syntax/Semantics distinction Models and entailment Soundness and completeness of inference procedures Syntax and semantics of propositional logic Validity and satisfiability (WalkSAT) Syntax and semantics of first-order logic Objects, relations and functions in FOL Common mistakes to avoid with quantifiers from FOL lecture (Universal and existential instantiation, and the relationship of the resulting knowledge bases to the original) (Unification) CIS 391 - Intro to AI
Logic: Skills you should have Is proposition P entailed by a knowledge base? Model-checking by hand Write out a truth table Exercises using important rules to know – e.g. de Morgan’s laws Converting statements into Conjunctive Normal Form Resolution Refutation Forward and backward chaining Representing a problem in propositional (or first-order) logic (Unification) CIS 391 - Intro to AI