Scribd
Upload
76 views

Mini-Test 1 2022

Uploaded by

zzy1211872190
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
76 views

Mini-Test 1 2022

Uploaded by

zzy1211872190
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

MBML 42186

Technical University of Denmark

SPRING, 2022
Campus: Lyngby

MODEL-BASED MACHINE LEARNING

Mini-test 1

(Time allowed: 75 MINUTES)

Name:
Student ID:

Part 1

Consider the following probabilistic graphical model (PGM):

z1 z2 z3

y1 y2 y3

u
–2– MBML 42186

1. (10%) Write the factorization of the joint distribution according to the PGM described above.

2. (25%) Formulate the expression for p(z3 |z2 , u) according to the factorization that you have written above.
Don’t forget to simply your expression to eliminate unnecessary factors.
–3– MBML 42186

3. (20%) Prove that:

(a) u ⊥
⊥ z3 |{y3 , x}
(b) x ⊥
⊥ y1 |{z1 , y2 , y3 }
–4– MBML 42186

Part 2

Consider the following generative process that models observed count data {yi,n }N
n=1 for each individual i
i

using a hierarchical modelling approach:

1. Draw µ ∼ N (0, 1)
2. Draw v ∼ Inverse-Gamma(1, 1)

3. For each individual i ∈ {1, . . . , I}:


(a) Draw individual-specific (log-)rate λi ∼ N (µ, v)
(b) For each observation n ∈ {1, . . . , Ni } of the individual i:
i. Draw observed count yi,n ∼ Poisson(exp(λi ))

Make sure that you understand it well.

1. (10%) Write the factorization of the joint distribution according to the model described above.

2. (15%) Draw the PGM representation corresponding to the generative process described above.
–5– MBML 42186

3. (20%) Suppose now that you also have access to covariates xi that describe characteristics of the each
individual i that could help explain his/her latent log-rate λi . You may assume that the relationship between
the mean of λi and xi is linear in parameters. Write the generative process corresponding to this enhanced
model. If you make any assumptions about the model when writing the new generative process, then make
them explicit and write them down.

You might also like