Machine Learning

Machine Learning 10-401, Spring 2018

Introduction, Admin, Course Overview
kassir
Lecture 1, 01/17/ 2018

Maria-Florina (Nina) Balcan

 Machine Learning
Image Classification

Document Categorization

Speech Recognition Protein Classification Spam Detection

Branch Prediction Fraud Detection Natural Language Processing

Playing Games Computational Advertising

2
Machine Learning is Changing the World
“Machine learning is the hot new thing”
(John Hennessy, President, Stanford)

“A breakthrough in machine learning would be worth ten

Microsofts” (Bill Gates, Microsoft)

“Web rankings today are mostly a matter of machine learning”

(Prabhakar Raghavan, VP Engineering at Google)
The COOLEST TOPIC IN SCIENCE
• “A breakthrough in machine learning would be worth
ten Microsofts” (Bill Gates, Chairman, Microsoft)
• “Machine learning is the next Internet”
(Tony Tether, Director, DARPA)
• Machine learning is the hot new thing”
(John Hennessy, President, Stanford)
• “Web rankings today are mostly a matter of machine learning”
(Prabhakar Raghavan, Dir. Research, Yahoo)
• “Machine learning is going to result in a real revolution”
(Greg Papadopoulos, CTO, Sun)
• “Machine learning is today’s discontinuity”
(Jerry Yang, CEO, Yahoo)
This course: introduction to machine learning.

• Cover (some of) the most commonly used machine learning

paradigms and algorithms.

• Sufficient amount of details on their mechanisms: explain

why they work, not only how to use them.

• Applications.
What is Machine Learning?
Examples of important machine learning paradigms.
Supervised Classification

from data to discrete classes

Supervised Classification. Example: Spam Detection
Decide which emails are spam and which are important.
Supervised classification
Not spam spam

Goal: use emails seen so far to produce good prediction rule for
future data.
 Supervised Classification. Example: Spam Detection
Represent each message by features. (e.g., keywords, spelling, etc.)

example label

Reasonable RULES:
+ + -
+ -
+
Predict SPAM if unknown AND (money OR pills)
-
- -
Predict SPAM if 2money + 3pills –5 known > 0 -
Linearly separable
Supervised Classification. Example: Image classification

• Handwritten digit recognition

(convert hand-written digits to
characters 0..9)

• Face Detection and Recognition

 Supervised Classification. Many other examples
• Weather prediction

• Medicine:
– diagnose a disease
• input: from symptoms, lab measurements, test results, DNA tests, …
• output: one of set of possible diseases, or “none of the above”
• examples: audiology, thyroid cancer, diabetes, …
– or: response to chemo drug X
– or: will patient be re-admitted soon?

• Computational Economics:
– predict if a stock will rise or fall
– predict if a user will click on an ad or not
• in order to decide which ad to show
 Regression. Predicting a numeric value

Stock market

Weather prediction

Temperature
72° F

Predict the temperature at any given location

 Other Machine Learning Paradigm
Clustering: discovering structure in data (only unlabeled data)
• E.g, cluster users of social networks by interest (community detection).
Facebook Twitter
network Network

Semi-Supervised Learning: learning with labeled & unlabeled data

Active Learning: learns pick informative examples to be labeled

Reinforcement Learning (acommodates indirect or delayed feedback)

Dimensionality Reduction

Collaborative Filtering (Matrix Completion), …

Many communities relate to ML
Admin, Logistics, Grading
 Brief Overview
• Meeting Time: Mon, Wed, NSH 3002, 10:30 – 11:50
• Course Staff
• Instructors:
– Mohamad Mahdi(Nina)
Maria Florina Kassir ([email protected])
Balcan ([email protected])
• TAs:
– Kenneth Marino ([email protected])
– Colin White ([email protected])
– Nupur Chatterji ([email protected])
 Brief Overview
• Course Website
http://www.cs.cmu.edu/~ninamf/courses/401sp18

• See website for:

• Syllabus details
• All the lecture slides and homeworks
• Additional useful resources.
• Office hours
• Recitation sessions
• Grading policy
• Honesty policy
• Late homework policy
• Piazza pointers

• Will use Piazza for discussions.

Prerequisites. What do you need to know now?
• You should know how to do math and how to program:
– Calculus (multivariate)
– Probability/statistics
– Algorithms. Big O notation.
– Linear algebra (matrices and vectors)
– Programming:
• You will implement some of the algorithms and apply
them to datasets
• Assignments will be in Octave (play with that now if
you want; also recitation tomorrow)
• Octave is open-source software clone of Matlab.
• We may review these things but we will not teach them
 Source Materials
No textbook required. Will point to slides and freely available
online material.
Useful textbooks:
Machine Learning, Tom Mitchell, McGraw Hill, 1997.

Machine Learning: a Probabilistic Perspective,

K. Murphy, MIT Press, 2012

Pattern Recognition and Machine Learning

Christopher Bishop, Springer-Verlag 2006
 Grading
• 40% for homeworks. There are 5 and you can drop 1.
• 20% for midterm [March 7]
• 20% for final [May 2nd]
• 15% for project
• 5% for class participation.
• Piazza polls in class: bring a laptop or a phone
• Homework 0: background hwk, out today [get full credit if you turn it in]

• Homeworks 1 to 4
– Theory/math handouts
– Programming exercises; applying/evaluating existing learners
– Late assignments:
• Up to 50% credit if it’s less than 48 hrs late
• You can drop your lowest assignment grade

• Projects: conduct a small experiment or read a couple of papers and present the main
ideas or work on a small theoretical question.
• Project presentations: April 23 and April 25
 Collaboration policy (see syllabus)
• Discussion of anything is ok…
• …but the goal should be to understand better, not
save work.

• So:
– no notes of the discussion are allowed…the only
thing you can take away is whatever’s in your
brain.
– you should acknowledge who you got help
from/did help in your homework
 Brief Overview
• Meeting Time: Mon, Wed, NSH 3002, 10:30 – 11:50
• Course Staff
• Instructors:
–- Mohamad Mahdi Kassir
Maria Florina (Nina)( [email protected]).
Balcan ([email protected])
• TAs:
– Kenneth Marino ([email protected])
– Colin White ([email protected])
– Nupur Chatterji ([email protected])
Maria-Florina Balcan: Nina
• Foundations for Modern Machine Learning
• E.g., interactive, semi-supervised, distributed, life-long learning

• Connections between learning & other fields (algorithms,

algorithmic game theory)
Game Theory
Approx. Discrete
Algorithms Optimization
Machine
Learning
Control Theory Matroid
Theory Theory
Mechanism Design

• Program Committee Chair for ICML 2016, COLT 2014

Kenneth Marino: Kenny
• Incorporating knowledge into Computer Vision
• Incorporating knowledge graphs
• Learning from Wikipedia articles

• Deep Learning – non-traditional training and architectures

• Graph Networks
• Generative Models (VAEs and GANs)
Colin White
• 4th year PhD student advised by Nina Balcan
• Design and analysis of algorithms

• Theoretical foundations of machine learning

• Beyond worst-case analysis

Worst-case Average case Real-world, application-specific

Nupur Chatterji
• Senior is SCS (Undergrad)

• Minor in Machine Learning (and Economics)

• Intend to pursue ML in grad school

• Interested in the intersection between technology and healthcare

 Learning Decision Trees.
Supervised Classification.
Useful Readings:
• Mitchell, Chapter 3
• Bishop, Chapter 14.4

DT learning: Method for learning discrete-valued

target functions in which the function to be learned
is represented by a decision tree.
 Supervised Classification: Decision Tree Learning

Example: learn concept PlayTennis (i.e., decide whether our friend

will play tennis or not in a given day)

Simple Day Outlook Temperature Humidity Wind Play Tennis

Training
Data Set
example label
 Supervised Classification: Decision Tree Learning
• Each internal node: test one (discrete-valued) attribute Xi
• Each branch from a node: corresponds to one possible values for Xi
• Each leaf node: predict Y (or P(Y=1|x  leaf))

Example: A Decision tree for

f: <Outlook, Temperature, Humidity, Wind>  PlayTennis?
Day Outlook Temperature Humidity Wind Play Tennis

E.g., x=(Outlook=sunny, Temperature-Hot, Humidity=Normal,Wind=High), f(x)=Yes.

 Supervised Classification: Problem Setting
Input: Training labeled examples {(x(i),y(i))} of unknown target function f
Day Outlook Temperature Humidity Wind Play Tennis

• Examples described by their values on

some set of features or attributes
• E.g. 4 attributes: Humidity, Wind, Outlook, Temp
– e.g., <Humidity=High, Wind=weak, Outlook=rain, Temp=Mild>
• Set of possible instances X (a.k.a instance space)

• Unknown target function f : XY

• e.g., Y={0,1} label space
• e.g., 1 if we play tennis on this day, else 0
Output: Hypothesis h  H that (best) approximates target function f
• Set of function hypotheses H={ h | h : XY }
– each hypothesis h is a decision tree
 Supervised Classification: Decision Trees
Suppose X = <x1,… xn>
where xi are boolean-valued variables

How would you represent the following as DTs?

f(x) = x2 AND x5 ? f(x) = x2 OR x5

𝑥2 𝑥2
1 0 1 0

𝑥5 𝑓 = 𝑁𝑜 𝑓 = 𝑌𝑒𝑠 𝑥5
1 0 1 0

𝑓 = 𝑌𝑒𝑠 𝑓 = 𝑁𝑜 𝑓 = 𝑌𝑒𝑠 𝑓 = 𝑁𝑜

Hwk: How would you represent X2 X5  X3X4(X1) ?

 Supervised Classification: Problem Setting
Input: Training labeled examples {(x(i),y(i))} of unknown target function f
Day Outlook Temperature Humidity Wind Play Tennis

• Examples described by their values on

some set of features or attributes
• E.g. 4 attributes: Humidity, Wind, Outlook, Temp
– e.g., <Humidity=High, Wind=weak, Outlook=rain, Temp=Mild>
• Set of possible instances X (a.k.a instance space)

• Unknown target function f : XY

• e.g., Y={0,1} label space
• e.g., 1 if we play tennis on this day, else 0
Output: Hypothesis h  H that (best) approximates target function f
• Set of function hypotheses H={ h | h : XY }
– each hypothesis h is a decision tree
Core Aspects in Decision Tree & Supervised Learning

How to automatically find a good hypothesis for training data?

• This is an algorithmic question, the main topic of computer science

When do we generalize and do well on unseen data?

• Learning theory quantifies ability to generalize as a function of the
amount of training data and the hypothesis space
• Occam’s razor: use the simplest hypothesis consistent with data!

Fewer short hypotheses than long ones

• a short hypothesis that fits the data is less likely to be a statistical coincidence
• highly probable that a sufficiently complex hypothesis will fit the data
Core Aspects in Decision Tree & Supervised Learning

How to automatically find a good hypothesis for training data?

• This is an algorithmic question, the main topic of computer science

When do we generalize and do well on unseen data?

• Occam’s razor: use the simplest hypothesis consistent with data!

• Decision trees: if we were able to find a small decision tree that

explains data well, then good generalization guarantees.
• NP-hard [Hyafil-Rivest’76]: unlikely to have a poly time algorithm

• Very nice practical heuristics; top down algorithms, e.g, ID3

 Top-Down Induction of Decision Trees [ID3, C4.5, Quinlan]

ID3: Natural greedy approach to growing a decision tree top-down (from the
root to the leaves by repeatedly replacing an existing leaf with an internal node.).

Algorithm: Day Outlook Temperature Humidity Wind Play

Tennis

• Pick “best” attribute to split at the root based

on training data.
• Recurse on children that are impure (e.g, have
both Yes and No).
Humidity
Outlook
Temp
Wind

Temperature Humidity Wind

Weak
High Normal
Strong Day Outlook Temperature Humidity Wind Play Tennis
Day Outlook Play Tennis
D1 Sunny Hot High Weak No
Sunny
Cool Overcast
Mild Rain
Hot D4 Rain Mild High Weak Yes
D2 Sunny Hot High Strong No D5 Rain Cool Normal Weak Yes
D8 Sunny Mild High Weak No D6 Rain Cool Normal Strong No
D9 Sunny Cool Normal Weak Yes D10 Rain Mild Normal Weak Yes
D11 Sunny Mild Normal Strong Yes D14 Rain Mild High Strong No

Humidity Yes Wind

High Normal Strong Weak

No Yes No Yes