Ocean Critters

1
Ocean Critters

• Dyana Vause
• Artificial Intelligence
• Fall 2005
• Dr. Lacher

2
A Working Platform

• IBM Thinkpad - Pentium 4
• Windows XP Professional
• GCL (GNU Common LISP) 2.6.1 for Windows
• Free
• CLtL compliant, ANSI compliance is its goal
• GNU Emacs for Windows
• Free
• Has a Common LISP mode
• Can have multiple windows
• Run, debug LISP inside Emacs

3
Where to Get It

• Did I mention its free?
• GCL - http//www.gnu.org/software/gcl/gcl.html
• Emacs - http//www.gnu.org/software/emacs/
• Good directions - http//www.cs.utexas.edu/users/n
  ovak/gclwin.html

4
My Working Platform

• Im old school, so
• Run the LISP executable directly
• Still free and almost ANSI compliant!
• Notepad for editing
• Free if you already have Windows
• Hey, I only use vi on Unix

5
The Expert System

• What it does - takes a set of rules, asks the
  user questions to determine which rules apply,
  uses a depth-first search to find the goal, and
  gives the user the possible answers with
  corresponding certainty factors
• Production system - has a set of rules, working
  memory, and the recognize-act cycle
• Unification algorithm for matching
• Depth first search
• Goal driven, backward chaining
• The rules do not support OR, or NOT, and so we
  are unable to say, for example, bald(x) -gt not
  hairy(x). The shell will ask both.

6
Borrowed Code

• Expert system code is borrowed from Lugers shell
  in Chapter 16
• unification.lsp
• streamfuncs.lsp
• expertshell.lsp
• No porting necessary and it worked on the first
  try!

7
Code

• unification.lsp
• recursive algorithm that performs matching /
  substitution on two expressions
• returns either failed, or the substitution list
  augmented with those bindings needed for a match
• expertsystem.lsp
• implements the shell, asks questions
• streamfuncs.lsp
• Implements the basic stream handling operations
  with delayed evaluation
• filters bindings that do not allow the goal to be
  satisfied
• oceanCritters.lsp
• my rules for determining some common critters
  found in or near the ocean

8
Certainty Factors

• Each rule has an associated certainty factor (cf)
• The cf of a solution is the product of all cfs of
  rules used
• Substitutions with a cf of less than 0.2 are
  pruned
• for example,
• (SKIN CRITTER FEATHERED) gt y
• (BREATHES CRITTER AIR) gty
• (COLOR CRITTER BLACKANDWHITE) gt y
• (FLIES CRITTER)gtn
• (SKIN CRITTER RUBBERY)gtn
• (SKIN CRITTER SCALY)gtn
• (SKIN CRITTER SHELLY)gtn
• (KIND CRITTER PENGUIN) cf 0.64
• notice that 0.8 0.8 0.64

9
How To Use the Expert Shell

• (load expertshell.lsp), (load
  unification.lsp), etc
• (lisp-shell)
• ( kind critter ( var x ) )
• answer the questions!

10
Expert Topic Ocean Critters

• skin(scaly) /\ breathes(water) -gt
  fish(x) .9
• skin(shelly) /\ breathes(water) -gt
  crustacean(x) .6
• skin(feathered) /\ breathes(air) -gt
  bird(x) .8
• skin(rubbery) /\ spinalColumnless(x) /\
  breathes(water)-gt invertebrate(x) .8
• skin(rubbery) /\ breathes(water) -gt
  shark(x) .9
• skin(rubbery) /\ breathes(air) -gt
  cetacean(x) .9
• bird(x) /\ color(blackAndWhite)
  -gt penguin .8
• bird(x) /\ flies(x)
  /\ color (white) -gt seagull .7
• bird(x) /\ flies(x)
  /\ color(brown) -gt pelican .9
• cetacean(x) /\ size(big) -gt
  whale .8
• cetacean(x) /\ size(medium) -gt
  dolphin .8
• fish(x) /\ bony(x)
  /\ snaky(x) -gt eel .9
• shark(x) /\ eats(plankton) -gt
  whaleshark .6
• shark(x) /\ eats(people) -gt
  greatWhite .9
• crustacean(x) /\ eats(garbage) /\
  size(tiny) -gt shrimp .8
• crustacean(x) /\ eats(garbage) /\
  size(small) -gt crab .7
• crustacean(x) /\ eats(plankton) -gt
  barnacle .6
• invertebrate(x) /\ has(feet) -gt
  squid .7
• invertebrate(x) /\ has(arms) -gt
  octopus .9

11
Demo!