Download Introduction to AI - Florida Tech Department of Computer Sciences

Introduction to Artificial Intelligence
(AI)
Abstract:
What is AI?
What happened within AI?
Going deeper: What really is AI?
Where AI happens?
What will you learn?
How will this course go?
Prerequisites
4/29/2017
(C) Debasis Mitra
Introduction to AI:
What is AI?
AI <is> Computer Science!
Whatever is at the FRONTIER of computer science is AI!
It is really about NEW challenges
that is CURRENTLY beyond computing
Chasing elusive goals!
4/29/2017
(C) Debasis Mitra
List of AI problems in text on slide 25:
http://cs.fit.edu/~dmitra/ArtInt/TextSlides/chapter01.pdf
As soon as “it” is achieved –
“it” is no longer AI!
AI is really a perception!
4/29/2017
(C) Debasis Mitra
Introduction to AI:
Historical Journey
Early Computers = Calculator plus stored program
• The Dream: Turing test, 1950
• Look Ma, No Hands! 1952-69
– Samuel’s checkers program: game playing
– Newell-Simon’s Logic Theorist: proving math theorems
– Geltner’s Geometry Engine: geometry theorems
• “AI” 1956 Dartmouth meeting
• Knowledge is power: Knowledge-based Systems, 1969-80
–
–
–
–
4/29/2017
Mycin, Dendral, and other expert systems
Expert systems industry boom, 1980-90
Fifth generation computing system (Japan)
MCC (US)
(C) Debasis Mitra
Introduction to AI:
Historical Journey
• Large knowledge bases, 1990-present
–
–
MCC Cyc project / IBM Watson
DARPA / SRI International /Apple Iphone Siri
• Machine learning
– Handwriting recognition, Speech recognition, by HMM 2000-10
– Big data, by CNN, 2010+
List of AI problems in text on slide 10:
http://cs.fit.edu/~dmitra/ArtInt/TextSlides/chapter01.pdf
4/29/2017
(C) Debasis Mitra
Introduction to AI:
What really is it?
• Computers manipulate numbers (hardware)
• Computers can map symbols internally to numbers (Unicode)
• First AI question: Can programs manipulate symbols?
– Symbolic AI: Expert systems
• Second AI question: Is symbol manipulation enough?
– No! Human intelligence is not just symbol manipulation
– And it is not efficient either as was thought of initially
• Third AI question: Should we go back to number manipulations
– Optimization: Machine learning
– …
• Future of AI: Numbers + Symbols!
– Data types: RealOrdinalPartial orderCategorical
4/29/2017
(C) Debasis Mitra
Introduction to AI:
Course content
• <<It is only algorithms, dummy!>> No “Intelligence”
– Because, there is no definition of intelligence
1. Search Algorithms
2. Constraint Reasoning
Mid term 1: 20%
3. Automated Reasoning
4. Probabilistic Reasoning
Mid term 2: 25%
5. Machine Learning
Project: 20%
6. Advanced topics
– CV, Robotics, NLP
4/29/2017
Final: 30%
(C) Debasis Mitra
Introduction to AI:
Course content: Samples
1. Search Algorithms
1.
2.
3.
Blind search
Guided search
Optimization
2. Constraint Reasoning
1.
2.
3.
Consistency levels
Back-jumping
Randomized algorithms
3. Automated Reasoning
1.
2.
3.
Logic
First order logic
Knowledge-base
4/29/2017
(C) Debasis Mitra
Introduction to AI:
Course content: Samples
4. Probabilistic Reasoning
4.
5.
6.
Bayesian network
Hidden Markov Modeling
Fuzzy logic
5. Machine Learning
5.
6.
7.
8.
Decision tree
Clustering
Training with samples
Neural networks
6. Advanced topics
–
–
–
–
Planning
Computer Vision
Robotics
Natural Language Processing
4/29/2017
(C) Debasis Mitra
Three levels of learning:
• Tourist-level: definition-concepts
• Deeper level: Solving exercises
• Expert level: Coding
• Industry! Frontiers
Introduction to AI: Where AI Happens
– Google, Amazon, Facebook, … start-ups, …
• Academic AI is far from real life, but
–
–
–
–
Challenges are addressed by new algorithms here
“Industry is 10 years behind academics,” a past PhD at Amazon now
AIMA: our Text by Russell-Norvig
My own humble interest: AI in Science and Engineering
• Conferences
–
–
–
–
AAAI: Association for Advancement (used to be American  ) of AI
IJCAI: International Joint Conference on AI
IEA/AIE: Applied AI
-…
• Journals
–
–
–
–
4/29/2017
AI journal
Journal of AI Research, online
Applied AI
…
(C) Debasis Mitra
Introduction to AI:
What can you do with this course?
• Some companies prefer students with AI background
• Current boom in Data Science (a new name for Data Mining)
• Helps in other advanced courses
–
–
–
–
–
Machine learning
Computer Vision
Scientific computing
Computational Molecular Biology
Complex network
• Research in/out our department
• Specialized algorithms / concepts
4/29/2017
(C) Debasis Mitra
Introduction to AI:
Pre-Quiz (One sentence each, total 7 min)
• Name:
• Level (UG/Grad):
• Major:
1.
2.
3.
4.
5.
What did you know about AI (i.e. your current perception)?
If you were to develop an AI program, what would that be on?
Name 3 topics that will be covered in this course.
Name one of the popular AI software or application.
Why are you taking this course (honest answer please)?
4/29/2017
(C) Debasis Mitra
Introduction to AI:
Required Background
•
•
•
•
•
•
•
List-Queue-StackTree-Graph-Spanning treeDjikstraRecursion-Convert to Iteration
Combinatorics
BigO-Loop Complexity-NPProgramming skill in any high level language:
–
4/29/2017
I encourage you to learn Python, it is “hot” now
(C) Debasis Mitra
Introduction to AI: Required Background
List-Queue-Stack-
List:
(pointer/array)
Stack:
(data type: data+operation
4/29/2017
(C) Debasis Mitra
Introduction to AI: Required Background
List-Queue-StackQueue:
(First-in-First-Out)
Max/Min Heap / Priority-Queue:
(a binary tree to provide
Max/Min value quickly
O(log N), N nodes)
4/29/2017
(C) Debasis Mitra
Introduction to AI: Required Background
Tree-Graph-Spanning treeGraph (V, E):
• Directed/ Un• Labeled / Un• Weighted / Not-
4/29/2017
(C) Debasis Mitra
Introduction to AI: Required Background
Tree-Graph-Spanning treeSpanning tree:
(Subgraph without loop, but covers all nodes)
Many spanning trees possible
On weighted graph,
Minimum spanning tree= least total wt
4/29/2017
(C) Debasis Mitra
Introduction to AI: Required Background
Shortest-path finding
Single-source-shortest-path-to-all:
(Djikstra’s algorithm)
4/29/2017
(C) Debasis Mitra
Introduction to AI: Required Background
Recursion-Iteration
Fibonacci series computing:
f(n) = f(n-1) + f(n-2),
f(0)=f(1)=1
4/29/2017
(C) Debasis Mitra
Introduction to AI: Required Background
Recursion-Iteration
Iterative fibonacci(n)
fib(0) = fib(1) = 1;
for i=2 through n do
fib(i) = fib(i-1) + fib(i-2);
end for;
return fib(n).
Avoids repetition:
4/29/2017
(C) Debasis Mitra
Introduction to AI: Required Background
Combinatorics
Permutation-Combination-Factorial-PowerSet
Quiz Questions (do later):
1. 20 nodes, how many pair of nodes?
2. 20 nodes, how many triangles possible?
3. 20 nodes, how many subsets of nodes?
4. 20 letters, how many unique words of length 20 possible?
4/29/2017
(C) Debasis Mitra
Introduction to AI: Required Background
BigO-Loop Complexity-NP-
Iterative fibonacci(n)
fib(0) = fib(1) = 1;
for i=2 through n do
fib(i) = fib(i-1) + fib(i-2);
end for;
return fib(n).
What is O(?) here?
What is O(?) here?
for i in range (1, n)
for j in range (1, m)
for k in range (1, n)
-whatever-
Q5. What is O(?) here, in terms of n and m?
Q6. What if you replace range of k to (1, 100)
4/29/2017
No Name or Id
UG/Grad, Major, and GPA if you know
Introduction to AI: Required Background
QUIZ
Quiz Questions:
20 nodes, how many pair of nodes?
20 choose 2, def of ‘choose’, = 20! / ((20-2)! * 2!)
20 nodes, how many triangles possible?
20 choose 3, = 20! / ((20-3)! * 3!)
20 nodes, how subsets of nodes?
2^20, each element absent/present in the subset, true/false
20 letters, how many unique words possible for word of length 20?
unique lettering: 20!, first position 20 possibilities * next position 19 possibilities * …
not unique lettering: 20^20, each position 20 possibilities, 20*20* … 20-times
max size of word up to 20: 20 + 20^2 + 20^3 + …+20^20
no max size specified: infinity
for i in range (1, n)
for j in range (1, m)
for k in range (1, n)
-whatever5. What is O(?) here, in terms of n and m?: O(n2m)
6. What if you replace range of k to (1, 100): O(nm)
4/29/2017
(C) Debasis Mitra