COMPSCI 188 - 2018-10-11 - Bayes' Nets: Independence (D-Separation)

Рет қаралды 23,561

Күн бұрын

COMPSCI 188, LEC 001 - Fall 2018
COMPSCI 188, LEC 001 - Pieter Abbeel, Daniel Klein
Copyright @2018 UC Regents; all rights reserved
"Slides (from 2018): inst.eecs.berkeley.edu/~cs188...
Latest website: inst.eecs.berkeley.edu/~cs188
More resources: ai.berkeley.edu
00:00 Setup [no content]
02:02 Announcements [outdated]
03:12 Review: Probability
06:35 Review: BNs
12:11 Size of a BN
15:57 Review: Conditional Independence
18:01 BN Independence Assumptions
24:07 Independence in a BN
27:44 Causal Chain Triple
32:03 Common Cause Triple
35:16 Common Effect Triple
38:53 D-Separation
43:37 Break [no content]
45:51 D-Separation Example 1
50:32 D-Separation Example 2
58:06 D-Separation Example 3
1:00:49 BN Structure Implications
1:05:19 Topology Limits Distributions
1:12:42 Summary
1:14:54 End [no content]"

Пікірлер: 7

@XernPL 5 жыл бұрын

3:17 begin

@muhammadsayed6495 3 жыл бұрын

39:55 shouldn't it be: "Attempt 1: if two nodes are connected by an undirected path blocked by a shaded node, they are conditionally independent". instead of "not blocked by a shaded node" ? because if they are blocked by a shaded node they will be most probably conditionally independent (the 2 other cases except the V shape case)

@njln1730 4 жыл бұрын

45:51 break ends

@le0nz Жыл бұрын

Thanks homie!

@SaurabhKumar-lu7su 3 жыл бұрын

The variables separated by comma in conditional probability expression are very confusing. e.g. P(z|x,y) could mean P(z|(x,y)) or P((z|x),y). Or is there some standard convention here that I am missing?

@anglarnasandel4428 3 жыл бұрын

The first is correct, as the second doesn't really make sense as far as I know. "the probability of z given x, also y must be true but for that one we don't know x" isn't really a thing that happens similarly P(z, x | y) would always mean P((z, x) | y) as opposed to P(z, (x|y)) I guess if it helps you can think of "," as having precedence over "|" or more accurately you could say the second is a separator between query and evidence (condition); you can't have more query variables after the separator.