19
Midterm Corrections & Quiz 3 Solutions Thur, July 14, 2016

Midterm Corrections Quiz 3 Solutions - Donald Bren School

  • Upload
    others

  • View
    4

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Midterm Corrections&

Quiz 3 Solutions

Thur, July 14, 2016

Page 2: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Midterm Problem 1.b.

Page 3: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Midterm Problem 1.b.

If we remove Attitude (A) and Participation (P) from the Bayes network figure, then the statement is correct.

T

Page 4: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Midterm Problem 1.b.

The correct statement should say:

Intelligence (I) and Grade (G) are conditionally independent given Test Scores (T) and Participation (P)

Page 5: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Midterm Problem 1.b.

Test Scores(T) and Attitude(A) are independent;However, Given Test Scores(T), Attitude(A) and Intelligence (I) are not independent!

Intuition Given Test Scores = Very Low, Intelligence and Attitude are not independent because: If we know that the student’s Intelligence = High, then we can infer that likely Attitude = Low.

Page 6: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Image from Mark Paskin Lecture on Graphical Models.[http://ai.stanford.edu/~paskin/gm-short-course/lec2.pdf]

Page 7: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Midterm Problem 1.b.

(Given)

Information flow.

Page 8: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Midterm Problem 6.j.

Page 9: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Midterm Problem 6.j.

In order to test conditional independence we have to test: P(A=t|B=t)P(C=t|B=t) =?= P(A=t, C=t|B=t) P(A=t|B=t)P(C=f|B=t) =?= P(A=t, C=f|B=t) P(A=t|B=f)P(C=t|B=f) =?= P(A=t, C=t|B=f) P(A=t|B=f)P(C=f|B=f) =?= P(A=t, C=f|B=f) P(A=f|B=t)P(C=t|B=t) =?= P(A=f, C=t|B=t) P(A=f|B=t)P(C=f|B=f) =?= P(A=f, C=f|B=t) P(A=f|B=f)P(C=t|B=f) =?= P(A=f, C=t|B=f) P(A=f|B=f)P(C=f|B=f) =?= P(A=f, C=f|B=f)

P(A=f|B=f)P(C=f|B=f) ≠ P(A=f, C=f|B=f)

Page 10: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Quiz 3 Problem 1

Page 11: Midterm Corrections Quiz 3 Solutions - Donald Bren School

When to do Goal-Test?

Lecture slides “Uninformed Search” Page 13

Page 12: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Goal-Test and Optimal Path

When to do Goal-Test? Guaranteed Optimal Path?

Breadth-first Search&Iterative Deepening Search

Before the child is pushed onto the queue.

No(Yes if cost is a non-decreasing function of depth)

Depth-first Search&Depth-limited Search

Before the child is pushed onto the queue.

No

Uniform Cost Search When the node is removed from the queue.

Yes

Greedy Best First Search Doesn’t matter because h(G) = 0 No

A* Search When the node is removed from the queue.

Yes if Admissible (tree search) or Consistent (graph search)

Page 13: Midterm Corrections Quiz 3 Solutions - Donald Bren School

Step Cost and Heuristic Value

● BFS, DFS, DLS, IDS does not consider step costs.

○ Costs are added up after finding a path.

● GBFS does not consider step costs, but considers heuristic value at each node.

● A* Search considers step costs and heuristic values.

Page 14: Midterm Corrections Quiz 3 Solutions - Donald Bren School
Page 15: Midterm Corrections Quiz 3 Solutions - Donald Bren School
Page 16: Midterm Corrections Quiz 3 Solutions - Donald Bren School
Page 17: Midterm Corrections Quiz 3 Solutions - Donald Bren School
Page 18: Midterm Corrections Quiz 3 Solutions - Donald Bren School
Page 19: Midterm Corrections Quiz 3 Solutions - Donald Bren School