6 hhkyrd - home | computer science at ubcschmidtm/courses/340-f17/t3.pdf · pp2`pb2r rl bp2"...

yi i i y

xi i i X

xij j i

d i

yi

p(yi|xi) ∝ p(xi|yi)p(yi)

we want to compare P(y=c|x_i) for different values of cand choose the maximum value

yi

p(yi|xi) ∝ p(xi|yi)p(yi)



= ...


xi yi



j=1

p(xij |yi)

P(xi1|yi)

P(xi2|yi)

P(xid|yi)

P(yi)

These are our parameters



= ...


xi yi



j=1

p(xij |yi)



= ...


xi yi



j=1

p(xij |yi)

our score for a given yi

We first need to compute our parameters

Prior: P(flu=N)= 3/6 =1/2

conditional: P(head=Y|flu=N)= 1/3

p( = N | = Y, = N, = Y ) ∝p( = Y | = N)p( = N | = N)p( =

Y | = N)p( = N) = 13 ∗ 2

3 ∗ 13 ∗ 1

2 = 0.0370

p( = Y | = Y, = N, = Y ) ∝p( = Y | = Y )p( = N | = Y )p( =

Y | = Y )p( = Y ) = 23 ∗ 1

3 ∗ 23 ∗ 1

2 = 0.0741

p(F |G) = 1 1%

p(F | ∼ G) = 0.01

p(G|F ) =?

F

G

∼ G

⇒

P (if depth is given)

P (if features are discrete)

Non-p

(the number of trees are fixed, but the depth usually varies with data) Non-p

P (k is given)

6 hhkyrd - home | computer science at ubcschmidtm/courses/340-f17/t3.pdf · pp2`pb2r rl bp2"...

Documents