1/17/2019 Lecture 1: Introduction
Foundations of
Data Analysis
Lecture 1: Introduction
[Link] 1/40
1/17/2019 Lecture 1: Introduction
[Link] 2/40
1/17/2019 Lecture 1: Introduction
Pierre-Simon Laplace (1749–1827)
[Link] 3/40
1/17/2019 Lecture 1: Introduction
Births in Paris
1745 - 1770
[Link] 4/40
1/17/2019 Lecture 1: Introduction
Births in Paris
1745 - 1770
251,527 Boys
[Link] 5/40
1/17/2019 Lecture 1: Introduction
Births in Paris
1745 - 1770
251,527 Boys 241,945 Girls
[Link] 6/40
1/17/2019 Lecture 1: Introduction
Are males born at a higher rate
than females?
251,527 Boys 241,945 Girls
[Link] 7/40
1/17/2019 Lecture 1: Introduction
Some possible statistics
251,527 Boys 241,945 Girls
[Link] 8/40
1/17/2019 Lecture 1: Introduction
Some possible statistics
251,527 Boys 241,945 Girls
Di erence: +9,582 Boys
[Link] 9/40
1/17/2019 Lecture 1: Introduction
Some possible statistics
251,527 Boys 241,945 Girls
Di erence: +9,582 Boys
Ratio: 104 Boys to 100
Girls
[Link] 10/40
1/17/2019 Lecture 1: Introduction
Some possible statistics
251,527 Boys 241,945 Girls
Di erence: +9,582 Boys
Ratio: 104 Boys to 100
Girls
Proportion: 50.97% Boys
[Link] 11/40
1/17/2019 Lecture 1: Introduction
Some possible visualizations
251,527 Boys 241,945 Girls
[Link] 12/40
1/17/2019 Lecture 1: Introduction
How did Laplace solve this?
[Link] 13/40
1/17/2019 Lecture 1: Introduction
How did Laplace solve this?
Conditional Probability that:
rate of boys, θ, is greater than girls,
given
observed data
[Link] 14/40
1/17/2019 Lecture 1: Introduction
Answer?
[Link] 15/40
1/17/2019 Lecture 1: Introduction
Answer?
P (θ > 0.5 ∣ data) = 1 − ϵ,
where ϵ ≈ 1 × 10
−42
.
[Link] 16/40
1/17/2019 Lecture 1: Introduction
What is probability?
[Link] 17/40
1/17/2019 Lecture 1: Introduction
What is probability?
De nition: Probability is the study of the
mathematical rules that govern random events.
[Link] 18/40
1/17/2019 Lecture 1: Introduction
But what is randomness?
[Link] 19/40
1/17/2019 Lecture 1: Introduction
But what is randomness?
Informally, a random event is an event where we
do not know the outcome without observing it.
[Link] 20/40
1/17/2019 Lecture 1: Introduction
But what is randomness?
Informally, a random event is an event where we
do not know the outcome without observing it.
Probability tells us what we can say about such
events, given our assumptions about the
possible outcomes.
[Link] 21/40
1/17/2019 Lecture 1: Introduction
What is statistics?
[Link] 22/40
1/17/2019 Lecture 1: Introduction
What is statistics?
De nition: Statistics is the application of
probability to the collection, analysis, and
description of random data.
[Link] 23/40
1/17/2019 Lecture 1: Introduction
Statistics is used to:
[Link] 24/40
1/17/2019 Lecture 1: Introduction
Statistics is used to:
Design experiments
[Link] 25/40
1/17/2019 Lecture 1: Introduction
Statistics is used to:
Design experiments
Summarize data
[Link] 26/40
1/17/2019 Lecture 1: Introduction
Statistics is used to:
Design experiments
Summarize data
Make conclusions about the
world
[Link] 27/40
1/17/2019 Lecture 1: Introduction
Statistics is used to:
Design experiments
Summarize data
Make conclusions about the
world
Explore complex data
[Link] 28/40
1/17/2019 Lecture 1: Introduction
What is machine learning?
[Link] 29/40
1/17/2019 Lecture 1: Introduction
What is machine learning?
De nition: Machine Learning builds statistical
models of data in order to recognize complex
patterns and to make decisions based on these
observations.
[Link] 30/40
Machine Learning is Everywhere?
Games Recommendation system
Assisted driving Cancer diagnosis
1/17/2019 Lecture 1: Introduction
[Link] 31/40
1/17/2019 Lecture 1: Introduction
Levels of data analysis
expertise
[Link] 32/40
1/17/2019 Lecture 1: Introduction
Levels of data analysis
expertise
0: What is data?
[Link] 33/40
1/17/2019 Lecture 1: Introduction
Levels of data analysis
expertise
0: What is data?
1: I know how to run data analysis software
[Link] 34/40
1/17/2019 Lecture 1: Introduction
Levels of data analysis
expertise
0: What is data?
1: I know how to run data analysis software
2: I understand the math behind the analysis
[Link] 35/40
1/17/2019 Lecture 1: Introduction
Levels of data analysis
expertise
0: What is data?
1: I know how to run data analysis software
2: I understand the math behind the analysis
3: I'm able to invent new data analysis methods
[Link] 36/40
1/17/2019 Lecture 1: Introduction
Why should you know the
mathematical foundations?
[Link] 37/40
1/17/2019 Lecture 1: Introduction
When machine learning goes wrong
from Goodfellow et al. ICLR 2015
[Link] 38/40
1/17/2019 Lecture 1: Introduction
When machine learning goes wrong
Panda (57.7% con dence)
from Goodfellow et al. ICLR 2015
[Link] 39/40
1/17/2019 Lecture 1: Introduction
When machine learning goes wrong
Panda (57.7% con dence) Gibbon (99.3% con dence)
from Goodfellow et al. ICLR 2015
[Link] 40/40