Final Review

1
Final Review

• CSC 533 Artificial Intelligence

2
Course Review

• Machine learning (25)
• Decision trees
• Current best hypothesis
• Version space learning
• Maximum likelihood and naïve bayes
• Artificial neural network
• Intelligent agent (10)
• Reflex, goal-based, utility-based, learning
• Search (35)
• Adversarial
• Uninformed
• Informed
• Logic (30)
• Propositional logic
• First order logic

3
Machine Learning

• Decision tree
• Current best hypothesis
• Version space
• Maximum likelihood
• Naïve bayes
• Artificial neural network

4
Decision Trees

• Entropy
• -p1logp1-p2logp2- -pn logpn
• Information gain
• Reduction in entropy by sorting on an attribute
• Entropy before sorting average entropy of all
  branches
• Sum of entropy of each branch weighted by
  percentage of examples sorted into that branch

5
Decision Trees (Cont.)

• Recursively build the tree top down
• Pick the attribute with the highest information
  gain as the root of the subtree
• Each branch to a subtree represent an attribute
  value
• Each leaf represents a prediction
• Prediction made by traversing the tree from root
  to the leaves
• Complete hypothesis space, incomplete search
• Occams razor

6
CBS

• Basic definitions
• Hypothesis
• Hypothesis space
• Syntactically/Semantically distinct
• General to specific partial ordering
• CBS
• Generalization of Find-S
• Generalize hypothesis to cover false negatives
  and specialize hypothesis to exclude false
  positives
• Backtracking from a dead end
• Non-deterministic solutions

7
Version Space

• Definition
• Consistent hypothesis
• Version space
• G boundary and S boundary
• Candidate-Elimination
• More compact representation of hypothesis space
  (S and G boundaries)
• Cannot handle noise (target will be eliminated)
• Optimal query to reduce the version space

8
Maximum Likelihood

• Bayes Theom
• Fundamental idea
• Compute probability of each hypotheses given data
• Predict according to the weighted sum of
  predictions made by all hypotheses

9
Maximum Likelihood (Cont.)
10
Naïve Bayes
11
ANN Perceptron
12
ANN Training Rules

• Perceptron works for linearly separable example
  space
• Delta works for non-linearly separable space

13
ANN Non-Linear Units

• Activation functions
• Sigmoid function continuous, differentialable

14
25 from Chap. 18.1-18.3, 19.1, 20.1-20.2, 20.5

• Example problems
• Definitions
• Find the S and G boundaries
• Compute a simple decision tree
• Train a simple bayesian classifier

15
Intelligent Agents

• Definitions
• Reflex agent
• Goal-based agent
• Utility-based agent
• Learning agent

16
10 from Chap 2

• Example problems
• Write a PEAS description of an agent
• Characterize an environment by the properties
  (accessible, deterministic, episodic, etc.)

17
Search

• Adversarial search
• Uninformed search
• Informed search

18
Adversarial Search

• Minimax algorithm (two player, multiplayer)

19
Adversarial Search (Cont.)

• Time and resource limit
• Cutoff search
• Evaluation function
• Equivalence class
• Alpha-beta pruning
• Pruning does not affect the final result
• Ordering matters

20
Problem Solving Agents

• Definitions
• Problem and goal formulation
• Problem types

21
Uninformed Search

• BFS (Breadth-first)
• DFS
• UCS
• DLS
• IDS
• Strategy evaluation

22
Informed Search

• BFS (Best-first)
• greedy
• A search
• Admissible, consistent, dominant
• Hill-climbing
• Iterative improvement
• Local optimal
• Simulated annealing
• Escape local optimal by allowing bad moves with
  decreasing probability

23
35 from Chap 3, 4, 6

• Example problems
• Definitions
• Alpha-beta pruning
• Generate search trees using BFS, DFS, IDS, etc.
• Draw the state space of a problem
• Derive or explain admissible heuristics
• A analysis

24
Logic

• Logic in general
• Entailment and model checking
• Propositional logic
• Syntax and semantics
• Logical proof
• Truth table
• Logical equivalence
• Validity and satisfiability
• Soundness and completeness
• Forward and backward chaining
• First order logic
• Syntax and semantics
• Universal quantification
• Existential quantification

25
30 from Chap 7 - 8

• Example problems
• Definitions
• Propositional logic vs. FOL
• Logical proof of equivalence, validity
• Converting between FOL and English
• Identify valid or invalid reasoning