Announcements

1
Announcements

• Todays Handouts
• Outline Class 4-5
• LISP Notes 1
• Web Site
• www.mil.ufl.edu/eel5840
• Software and Notes
• Reading Assignment
• Nilsson Chapter 3
• Written Assignment Reminder
• Homework 2 due Thurs. 9/4 in class

2
Todays Menu

• Architectures for the Implementation of Action
  Functions
• B. State Machines
• C. Artificial Neural Networks
• D. Subsumption Architecture
• Final Thoughts on Stimulus-Response (SR) Agents -
  Chapter 2
• Introduction to the AI Language LISP
• LISP
• Chapter 2 Basic LISP Primitives
• Chapter 3 Procedure Definition Binding
• Chapter 4 Predicates Conditionals

3
Perception and Action

• State Machines Implementation of Boolean
  (action) functions using a connected network of
  logical gates AND,OR,NOR, etc.)
• Networks Implementation of action functions
  using a connected network of threshold units or
  other elements that compute a nonlinear function
  of a weighted sum of their inputs. One such
  element is the threshold logic unit or TLU for
  short.

Boolean functions implementable by a TLU are
called linearly-separable functions. (A TLU
separates the space of input vectors into an
above-threshold response from below-threshold
response by a linear surface?called a hyperplane
in n dimensions.) Not all Boolean functions are
linearly separable?however a monomial or any
clause is linearly separable.
4
Perception and Action
Example Let fx1/x2x3 x1/x2x3 f ? TLU
0 0 0 0 0 0 0 0 1 0 1
0 0 1 0 0 -1 0 0 1 1 0 0
0 1 0 0 0 1 0 1 0 1 1
2 1 1 1 0 0 0 0 1 1 1
0 1 0
Example Let b-f4x1/x2 (s2s3)/(s4s5)(s2s3)/
s4/s5 s2 s3/s4/s5 b-f ? TLU s2 s3/s4/s5 b-f ?
TLU 0 0 0 0 0 0 0 1 0 0 0 1
1 1 0 0 0 1 0 -2 0 1 0 0 1 0
-1 0 0 0 1 0 0 -2 0 1 0 1 0
0 -1 0 0 0 1 1 0 -4 0 1 0 1 1
0 -3 0 0 1 0 0 1 1 1 1 1 0
0 1 2 1 0 1 0 1 0 -1 0 1
1 0 1 0 0 0 0 1 1 0 0 -1 0
1 1 1 0 0 0 0 0 1 1 1 0 -3 0
1 1 1 1 0 -2 0
5
Perception and Action

• The Subsumption Architecture
• An agents behavior is controlled by a number of
  behavior modules. Each module receives sensory
  information directly from the world. If the
  sensory inputs satisfy a precondition specific to
  that module, then a certain behavior, also
  specific to that module, is executed. One
  behavior module can subsume another.
• As contrasted with much other work in AI, these
  machines do not depend on complex internal
  representations of their environments or on
  reasoning about them.

6
Neural Networks
Continuous FA Classifications Predictions
Adjustable Model
Training Data
Training Algorithm

• Neural Networks (also known as Artificial Neural
  Networks or ANNs for short)
• You need this framework to model processes that
  cannot be represented as analytical models, e.g.,
  human actions, computer vision, non-linear
  control, the stock market...

7
The AI Language LISP

• LISP - LISt Processing Invented in the late 40s
  by John McCarthy at MIT on an IBM 709 computer.
• LISP is about symbolic processing, i.e., symbol
  manipulation is treating the binary quantities
  inside the computer like the words and sentences
  of a language. The words in LISP re called atoms.
  The sentences are called lists. Collectively
  atoms and lists are called symbolic expressions
  or s-expressions or SEX for short.
• Examples
• (arroyo (professor ece)
• (degree phd)
• (area (ce robotics) ) )
• (trip (gainesville tallahassee 150)
• (tallahassee perry 50)
• (perry gainesville 100) )

8
The AI Language LISP

• Uses of LISP
• Expert Problem Solvers
• Commonsense Reasoning
• Learning
• Natural Language Interfaces
• Education and Intelligent Support Systems
• Speech and Vision
• The premier symbolic processing language is
  Common LISP
• Myths
• LISP is slow
• LISP programs are big
• LISP is hard to learn
• LISP is hard to debug read because all those
  parentheses

9
The AI Language LISP

• Tutorial Introduction to LISP
• XLISP is available from stat.umn.edu by David
  Betz Luke Tierney
• We have the latest 16-bit 32-bit version on
  www.mil.ufl.edu/eel5840
• Install using winzip, pkunzip or in a fresh
  directory by running the self-extracting file.
• Go to the XLISP-STAT resources link to obtain
  additional information including a manual, the
  UNIX version and the MAC version.

Objects
Input
Atoms
Lists
Eval
Numeric
Alpha
10
LISP Lab 1

• Predicate Functions
• (atom sex) t if sex is an atom
• (null sex) t if sex is nil or ()
• (eq sex1 sex2) t if sex1sex2 (identical)
• (equal sex1 sex2) t if sex1sex2
• (zerop sex1) t if sex0
• (numberp sex) t if sex is a number
• (symbolp sex) t if sex is a symbolic atom
• (listp sex) t if sex is a list
• (member sex lis) nil if sex is not a member of lis

11
LISP Lab 1

• LIST Functions
• (car lis) returns the 1st sex of list lis,
  same as first
• (cdr lis) returns the list lis with the 1st
  sex removed, same as rest
• (car (cdr lis)) same as second, i.e., 2nd sex of
  the list lis
• (list sex1 sex2) returns the list (sex1 sex2)
• (cons sex1 lis) makes sex1 the 1st element of
  the list lis
• (append lis1 lis2) a new list with all the
  elements of lis1 followed
• by all the elements of lis2
• (defun fname (argument-list) lt(forms)gt)
• (pprint (function-lambda-expression fname))

12

• See Class 4-5 Notes
• LISP Tutorial 1
• The End!