Scribd
Upload
193 views

Aml Assignment 3 Bayesian Linear Regression PDF

This 3 sentence summary provides the essential information about the document: The document is a project guide for an assignment on Bayesian linear regression that will test skills in calculating Bayesian posteriors and building functions to calculate the parameters of Bayesian posteriors for Bayesian linear regression using housing data to predict house prices from living area and year built, with an expected completion time of 3 hours.

Uploaded by

Dong Hai Nguyen
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
193 views

Aml Assignment 3 Bayesian Linear Regression PDF

This 3 sentence summary provides the essential information about the document: The document is a project guide for an assignment on Bayesian linear regression that will test skills in calculating Bayesian posteriors and building functions to calculate the parameters of Bayesian posteriors for Bayesian linear regression using housing data to predict house prices from living area and year built, with an expected completion time of 3 hours.

Uploaded by

Dong Hai Nguyen
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

aml_assignment_3_bayesian_linear_regression

January 16, 2020

1 Bayesian Linear Regression

Author: Khal Makhoul, W.P.G.Peterson

1.1 Project Guide

• Section ??
• Section ??
• Section ??
• Section ??

## Project Overview

EXPECTED TIME 3 HRS This assignment will test your abilities in two different sections: the
Section ?? section will revisit Bayes’ formula and evaluate your ability to calculate simple Bayesian
posterior probabilities. The Section ?? section will ask you to build functions that calculate the
parameters of Bayesian posteriors for Bayesian Linear Regression.
In a separate notebook, there is a brief demonstration of pymc3. The pymc3 module is a stan-
dard for using Markov Chain Monte Carlo to estimate Bayesian derived distributions, in lieu of
calculating exact integrals.
One note on pymc3: it may be tricky to make function on your individual machine
Programming questions will include: - Calculation of Bayesian Posterior - Calculating the
w MAP - Estimating σ2 - Calculating Σ
Motivation: Bayesian regression allows us to quantify the uncertainty in our model building
/ the point estimates of our weights calculated in Least Squares Regression.
Objectives: This assignment will: - Test fundamental Bayesian knowledge, particulary in re-
gard to Linear Regression - Introduce pymc3
Problem: Once again we will be using housing data to predict house price using living area
and year built.
Data: Our data comes from Kaggle’s House Prices Dataset.

1
Imports:

In [1]: ### This cell imports the necessary modules and sets a few plotting parameters for disp

%matplotlib inline
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
plt.rcParams['figure.figsize'] = (20.0, 10.0)

### Introduction and Review.


Bayesian Regression comes with a different toolset than ordinary Linear Regression. In turn,
that toolset demands a slightly different mindset. We start with a short review to hightlight the
ways in which Bayesian thinking proceeds.
Consider a population whose age distribution is as follows:

Age group % of total population


≤ 35 25%
36 − 65 45%
≥ 66 30%

Say you know the following results of a study about YouTube viewing habits:

Age group % in this group that watch YouTube every day


≤ 35 90%
36 − 65 50%
≥ 66 10%

Prompt: If you know a user watches YouTube every day, what is the probability that they
are under 35?
We will start with a prior, then update that prior using the likelihood and the normalization
from Bayes’s formula. We define the following notation:

• A: YouTube watching habit


• B: Age
• A = 1: User watches YouTube every day
• A = 0: User does not watch YouTube every day
• B ≤ 35: User has age between 0 and 35
• 36 ≤ B ≤ 65: User has age between 36 and 65
• B ≥ 66: User has age greater than 65

The prior can be read from the first table:

P( B ≤ 35) = 0.25
We are looking to calculate the posterior probability:

P( B ≤ 35| A = 1)

You might also like