Scribd
Upload
31 views

CEE598 Fall 2016: Homework Assignment 6 (Revised On Nov 9) : 2 Prior 2 Prior

This document provides instructions for Homework Assignment 6 which is due on November 14, 2016. It includes two problems. Problem 1 asks students to show that the inverse-gamma distribution is the conjugate prior for the variance of a Normal distribution with known mean based on N data points. It also asks students to find the parameters of the posterior distribution. Problem 2 provides experimental retention data and asks students to (a) find the maximum likelihood estimates, (b) calculate the MAP estimates with uniform priors between 0 and 0.5, (c) calculate the MAP estimates with uniform priors between 0 and 1, and (d) discuss the results for a power model of forgetting.

Uploaded by

gongmaozhe
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
31 views

CEE598 Fall 2016: Homework Assignment 6 (Revised On Nov 9) : 2 Prior 2 Prior

This document provides instructions for Homework Assignment 6 which is due on November 14, 2016. It includes two problems. Problem 1 asks students to show that the inverse-gamma distribution is the conjugate prior for the variance of a Normal distribution with known mean based on N data points. It also asks students to find the parameters of the posterior distribution. Problem 2 provides experimental retention data and asks students to (a) find the maximum likelihood estimates, (b) calculate the MAP estimates with uniform priors between 0 and 0.5, (c) calculate the MAP estimates with uniform priors between 0 and 1, and (d) discuss the results for a power model of forgetting.

Uploaded by

gongmaozhe
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

CEE598 Fall 2016: Homework Assignment 6 (Revised on Nov 9)

Due on Monday Nov 14, 2016, 8:30 am


Instructor: Hadi Meidani ([email protected])

Note: Updates are highlighted below.


Problem 1. In the lecture, we saw an example of a conjugate prior of Normal distribution family, when
measurement data were considered to be drawn from a Normal distribution with known variance, 2 , but
2
unknown mean , where N (prior , prior
). We derived the parameters of the posterior distribution for
a single data point. Now consider a slightly different case, where instead of one data point, N measurement
data points are available, and they are still believed to be from a Normal distribution, but this time with
known mean, , but unknown variance, = 2 . Show that a conjugate prior for the unknown variance is the
inverse-gamma distribution, given by
b
(; a, b) (a+1) exp( ).

Also, find the expressions for the parameters of the posterior distribution, i.e. apost and bpost .
Problem 2. In psychological sciences, forgetting curve is a concept used to characterize the decline of memory
retention. The curve is plotted with x-axis being the number of days after a newly learned knowledge and
y-axis being the percentage of that knowledge retained up to that day. An actual experiment is conducted
and the observed proportion of correct recalls (C) and incorrect recalls (W) are recorded. The following
observation data are reported with n = 100 independent trials, where t = [1, 3, 6, 9, 12, 18], are the time
intervals and y = [0.94, 0.77, 0.40, 0.26, 0.24, 0.16 ] are the observed proportion of correct recalls (C), which
indicates retention. A widely used model to predict the retention is the power model given by
(C; w1 , w2 , t) = w1 tw2
where w1 and w2 are the two model parameters, and (C; w1 , w2 , t) is the model prediction of the probability
of correct recalls. Using the provided experimental data
(a) Find the Maximum Likelihood estimates for w1 and w2 .
(b) Consider the two priors for these two parameters to be w1 U (0, 0.5) and w2 U (0, 0.5). Calculate the
MAP estimates for the two parameters.
(c) Consider the two priors for these two parameters to be w1 U (0, 1) and w2 U (0, 1). Calculate the MAP
estimates for the two parameters.
(d) Discuss your results.
Hint: This problem is very similar to the coin flip example.

You might also like