Professional Documents
Culture Documents
Differences
Between VR
and AR
Augmented Marries personal experiences with related imagery and
simulations
reality is
Adds value to the real world experience
more
Enables superior decision-making with additional
personal superimposed data
and Looks for and provides the missing elements that can
complex accelerate decision-making!
Big Data are high-volume, high-velocity, and/or high-variety
STRUCTURED DATA
European
As of 2017, administrators Poor data
53% of Forbes 40% projected
could save across Poor data can
surveyed Bad data or growth in
more than businesses cost
companies poor data global data
€100 billion and the businesses
adopted Big quality costs generated per
($149 billion) government 20%–35% of
Data US businesses year vs. 5%
in operational costs the U.S. their
Analytics. $600 billion growth in
efficiency economy $3.1 operating
annually. global IT
Up From 15% improvements trillion dollars revenue.
spending.
in 2015 alone by using a year.
big data
Unstructured The key need in today’s digital
Data environment is to:
1. Reduce the dimensionality of big data
(aggregate)
Structured
Data 2. Transform and synthesize
3. Extract intelligence (usable
information)
Useful
Information 4. Analyze (understand relationships)
5. Value-added decision-making
Analytics
Decision-
Making
Big Data = Big Nothing, Unless…
Descriptive Analytics: Inference
Identify Draw from
goals sample
Start Population Sample
Make inferences
about population
Population Sample Statistics
parameters
Source: Bennett, Briggs, and Triola (2014), Statistical Reasoning for Everyday Life.
9
Inference: Causality Versus Correlation
No. of Eggs
No. of Hens
10
Correlation Coefficients… Correlation = 0
Dependent variable
• Correlation 30
25
• If two events (features or 20
measurement variables) 15
tend to occur together, they 10
5
are said to be correlated 0
• Example: Lighting and 0 2 4 6 8 10 12
thunder, weight and height Independent variable
• Correlation coefficient:
Measuring the strength of • Outliers
linear relationship of • An outlier that is consistent with the trend of
measurement variables the rest of the data can inflate the correlation
• An outlier that is not consistent with the rest
• r è +1: Positive correlation of the data can deflate the correlation
• r è -1: Negative correlation
• Simpson’s paradox
• r è 0: No correlation (?) • Groups combined inappropriately may mask
relationships
• The missing link (effect of a third variable) 11
Example: Death rate vs. speed limit
13
Lurking vs. confounding variables
• Confounding occurs in a study
A
when you are unable to
B Causation
distinguish among the effects of
two or more explanatory
variables
• Potential problem
A B
• Example: Testing the effectiveness
Common response:
of a new drug
A B C = Lurking variable • Design of experiments: Give
C placebo to patients in clinic A and
real drug to patients in clinic B
Correlation • Even if we find difference between
exists patients in clinics A and B, what
kind of conclusion we can draw?
A ? B • The difference is due to the new
C= Confounding drug or clinics?
variable
C
14
Why should we care about correlation?
• How to confirm causation?
• Use of randomized experiments
(treatment-control study)
• Use of observational (case-control)
•
•
Easily observable
No proof needed
study if experiments are not possible or
ethical
• Evidence of a possible causal
connection:
• Not easily observable • There is a reasonable explanation of
• Require proof cause and effect
• The connection happens under varying
conditions
• Potential confounding variables are
ruled out
• Has research potential
• Most useful to us • Example: Developing a new medicine
15
Where Big Data and AI meet for a movie…
https://www.youtube.com/watch?v=ImpV70uLxyw
Driverless Cars
https://www.youtube.com/watch?v=ftouPdU1-Bo
Credit Fraud
Detection
https://www.youtube.com/watch?v=AuwayMjvuT0
Autonomous Storage and Handling
https://www.youtube.com/watch?v=UtBa9yVZBJM
Defining AI
• For e.g. Science which
makes machines Systems mimicking Rational thinking
perform real-time tasks human reasoning systems
that require higher
intelligence,
independent of human
intervention and
decision-making
• AI is the study of formal
AI
structures and problem
solving methods, that
aims to equip
computing machines Systems mimicking Rational acting
with capabilities to
solve problems, that
are comparable to
human behavior systems
human beings
Using formal rules to draw valid conclusions
Philosophy What does knowledge truly mean?
How can we convert knowledge into action?
Creating formal rules to draw valid conclusions
Measures and computation
Mathematics
Probability and inference
Fuzzy logic
Structures in AI
Behavioral conditioning
Psychology Cognitive psychology
Cognitive science
Common Tasks in AI