class website: http://csis.pace.edu/~benjamin/teaching/cs827/

Qualifying exam is not optional (if you completed exams requirement, it will just determine your course grade):

- Problems of the qualifying exam will be similar to those of the assignments.
- There will be Bayes rule in the exam.
- For the complete list of exam topics that you will be HELD RESPONSIBLE for, see: http://csis.pace.edu/~benjamin/teaching/cs827/topics.html

- Meeting schedule:
- 3 more flexible and upon request meetings.
- online video meetings is an option too.

- Homework submission:
- by email (anytime).

Overview on type of problems (and algorithms) in AI:

- In general, AI deals with ill-structured problems.
- Search algorithms (initial state --> describe the problem --> explore the possible solutions 'usually exponential' ).
- DFS, BFS, A*, iterative depening ... etc.
- You are expected to KNOW these basic algorithms in the Qualifying Exam.

**Describing search algorithms:**

*Global Search algorithms*

DFS:

- backtracking (conceptually or metaphor), but in reality it's implemented using datastructure not really backtrack.
- Not complete, but more space efficient (linear space)

BFS:

- Complete, but bad space complexity (exponential)

Iterative deepining:

- Combining DFS and BFS.
- Will be on the exam.

How to determine which approach to use?

Heuristic approach is a guess (e.g. how far you are from your goal) based on background knowledge about the problem domain.

**Greedy approach**employees the heuristic approach (using evaluation function).

Hint: all course units (except 5 and 6) are worked out with examples at: http://csis.pace.edu/~benjamin/teaching/cs627

- A* algorithm:
- the most used search algorithm in AI.

**Games algorihtms**

- Minimax:
- see: http://csis.pace.edu/~benjamin/teaching/cs627/webfiles/games/games3.html

*Local Search Algorithm*

- Random-Restart (recent and popular) algorithm (you must read it).
- Easy to implement and less cost.

- start at a randomly selected spot
- search locally (tune) if no result: terminates and go to 1

**Constraint Satisfaction**

Reviewing Constraint Satisfaction Problems:

```
- ex: place n queen in NxN chess board with no queen attacking the other.
- See: Example: Cryptarithmetic (chapter5.pdf page.9) and http://csis.pace.edu/~benjamin/teaching/cs627/webfiles/csp/csp.15b.html
- See: http://csis.pace.edu/~benjamin/teaching/cs627/webfiles/csp/csp.1.html
```

**Logic**

- Propositional Logic:
- expressing facts.

- Syllogism:
- symbolic way to conclude facts.

- Predicate logic:
- boolean value functions.
- e.g. sunny(Friday) = True

vacuously true.

See exercise 8.9:

`- in exam , we will have similar questions.`

**This is the Final Meeting of this class**

Qualifying exam preminlinary date: Wed. May, 27 2015

Discussing homework answers

- exercise 13.13

in response to a question, discussing Entropy (Information theory) and decision tree:

- see: http://csis.pace.edu/~benjamin/teaching/cs627/webfiles/learn/learn.13.html
- Qualifying exam will have something similar to this.

In response to a question, discussing probability and bayes:

- see: http://csis.pace.edu/~benjamin/teaching/cs627/webfiles/uncert/uncert.5.html