Week 3- BN

Question 1

BN

Answer 1

“Humans and AI make decisions based on BELIEFS about STATES IN THE WORLD (can be anything)
Beliefs = probability
P between 0 and 1
Not just p of current event but info that could affect that p

Question 2

Random Variable and CPT

Answer 2

RV: has a set of values that it can take on. Called its domain. Each node has this.
Nodes are connected by directed edges & each node has a CPT

CPT: probability of the nodes values given the values of its parents (if any)

Question 3

Indirect vs Direct dependencies in BN

Answer 3

The link between nodes represents this

Question 4

Querying a Node

Answer 4

To obtain the probability that it has a certain value

Ex. The probability of normal late being true

Question 5

Observing/Clamping a Node

Answer 5

Setting a node to a specific value
Ex. what is the probability of the node having this specific value
A belief propagation: When a node is observed

Question 6

Independent vs Dependent

Answer 6

Independent: if a given node Y is independent of node X, no information is transmitted throughout the network

Dependent: If Node Y and Node X are dependent, we get a different answer when we query Y after evidence on X is introduced

Question 7

Why is it important tot know when two nodes are independent

Answer 7

If they are independent they are not related. Evidence on one does not effect the other

Question 8

Determining if BN nodes are independent of each other: 3 things

Answer 8

Either: Serial connections
Diverging connections
Converging connections

• Presence/Absence of evidence
  AND
• Conditional Probability Table: P has to equal to 1

Question 9

Serial Connections - BN

Answer 9

Does not matter the order that you introduce evidence here

Question 10

Diverging Connection

Answer 10

If there is Hard evidence on parent node A: B and C are independent

If no hard evidence on A. B and C are dependent and evidence can flow between them.

Question 11

Converging Connection

Answer 11

(think of this flipped around)
If no evidence on A or A’s children, B and C are independent (they’re the parent nodes?)

If we have hard evidence on A or its children, B and C are dependent

Question 12

Bayesian Networks: Types of Inference

Answer 12

Diagnostic
Predictive
Mixed
Inter causal

Question 13

Diagnostic Reasoning BN

Answer 13

Reasoning occurs in the opposite direction (bottom up)

From EFFECTS to CAUSES

Question 14

Predictive Reasoning

Answer 14

Top to bottom

CAUSES to BELIEFS

Question 15

Inter causal Reasoning

Answer 15

??