AIMA JAVA Notes By Ravi(magesmail@yahoo.com)

AIMA JAVA Notes By Ravi(magesmail@yahoo.com) and Ciaran (ctjoreilly@gmail.com).

#summary some notes

= Introduction =

The latest (and ever evolving) code can be found at http://code.google.com/p/aima-java/. if you notice a bug please try checking out the latest version from the svn repository to see if it persists.

Current release is 0.94:<br>

Current release is 0.95:<br>

This is our first release containing GUIs (thanks to Ruediger Lunde):<br>

- aima.gui.applications.VacuumAppDemo<br>

Provides a demo of the different agents described in Chapter 2 and 3

for tackling the Vacuum World.<br>

- aima.gui.applications.search.map.RoutePlanningAgentAppDemo<br>

Provides a demo of the different agents/search algorithms described

in Chapters 3 and 4, for tackling route planning tasks within

simplified Map environments.<br>

- aima.gui.framework.SimpleAgentAppDemo<br>

Provides a basic example of how to create your own Agent based

demonstrations based on the provided framework.<br>

<br>

This will also be our last full release based on the 2nd edition of AIMA.

We are currently in the planning phases to re-organize this project based on the 3rd edition of AIMA, which should be available soon.

Previous release is 0.94:<br>

This is a patch release for the FOL Logic and includes the following fixes:<br>

- Fixed subtle defect in Model Elimination inference algorithm, which caused it to miss portions of the search space.<br>

- Improved the performance of both theorem provers, in particular added support for forward and backward subsumption elimination, which improves significantly the performance and use of the OTTER Like theorem prover.<br>

- a completion of the First Order Logic concepts from Chapter 9.<br>

- the addition of the LRTA Agent from Chapter 4.<br>

Note: If running the unite tests be sure to include the vm arguments:

Note: If running the unit tests be sure to include the vm arguments:

-Xms256m -Xmx1024m

as some of the First Order Logic algorithms (i.e. FOLTFMResolution) are

memory hungry.

Heuristics are now doubles (vs ints in the old version).

One minor change is that I've dropped the make file. Please use [http://ant.apache.org/ant ant]

==bug reports - acknowledgment ==

The following people sent in excellent comments and bug reports. Thank you!!!!

* Ali Tozan

* Carl Anderson, Senior Scientist, ArchimedesModel.com

* Don Cochrane from (?) University

* Mike Angelotti from Miami University

* Chad Carff ,University of Western Florida . EXCELLENT test cases . thank you .

* Dr .Eman El-Sheikh, Ph.D.,University of Western Florida

* Ravindra Guravannavar, Aztec Software,Bangalore

* Cameron Jenkins,University Of New Orleans

* Nils Knoblauch (Project Manager, Camline) - winner of the No Prize for the best bug report ! Thanks!

* Phil Snowberger, Artificial Intelligence and Robotics Laboratory,University of Notre Dame

==Bug Reports - acknowledgment ==

The following people sent in excellent comments and bug reports. Thank you!!!!<br>

* Ali Tozan<br>

* Carl Anderson, Senior Scientist, ArchimedesModel.com<br>

* Don Cochrane from (?) University<br>

* Mike Angelotti from Miami University<br>

* Chad Carff ,University of Western Florida . EXCELLENT test cases . thank you.<br>

* Dr .Eman El-Sheikh, Ph.D.,University of Western Florida<br>

* Ravindra Guravannavar, Aztec Software,Bangalore<br>

* Cameron Jenkins,University Of New Orleans<br>

* Nils Knoblauch (Project Manager, Camline) - winner of the No Prize for the best bug report ! Thanks!<br>

* Phil Snowberger, Artificial Intelligence and Robotics Laboratory,University of Notre Dame<br>

= Details =

==Build Instructions==

If you just want to use the classes, all you need to do is put the aima-java/build directory on your CLASSPATH.

if you want to rebuild from source, run the unit tests etc follow the instructions below.

# put [http://prdownloads.sourceforge.net/junit/junit3.8.1.zip?download junit 3.8.1 (note the version number)] on the classpath

# type 'ant'

I have included the eclipse.classpath and .projectfiles for those who use [http://www.eclipse.org eclipse] .

==Code Navigation==

# To understand how a particular feature works , FIRST look at the demo files.There are four main demo files SearchDemo , LogicDemo ,ProbabilityDemo and LearningDemo.

# If the Demo Files don't exist yet , look at the unit tests . they often cover much of how a particular feature works .

# If all else fails , write to me . Comprehensive documentation, both java doc and otherwise are in the pipeline , but will probably have to wait till I finish the code .

==Notes on Search==

To solve a problem with (non CSP )Search .

# you need to write four classes .

# a class that represents the Problem state .This class is independent of the framework and does NOT need to subclass anything . Let us, for the rest of these instruction, assume you are going to solve the NQueens problem . So in this step you need to write something like aima.search.nqueens.NQueensBoard .

# a subclass of aima.search.framework.GoalTest.This implements only a single function ---boolean isGoalState(Object state); The parameter state is an instance of the class you created in step 1-a above. For the NQueensProblem you would need to write something like aima.search.nqueens.NqueensBoardTest

# a subclass of aima.search.framework.SuccessorFunction .This generates a stream of Successors where a Successor is an object that represents an (action, resultantState) pair. In this release of the code the action is a String (something like "placeQueenAt4,4" and the resultant State is an instance of the class you create in step 1.a . An example is aima.search.nqueens.NQueensSuccessorFunction.

# If you need to do an informed search, you should create a fourth class which subclasses aima.search.framework.HeuristicFunction. This implements a single function int getHeuristicValue(Object state); keep in mind that the heuristic should DECREASE as the goal state comes nearer . For the NQueens problem, you need to write something like aima.search.nqueens.QueensToBePlacedHeuristic.

that is all you need to do (unless you plan to write a different search than is available in the code base ).

To actually search you need to

}

}

}}}

==Search Inheritance Trees ==

there are two inheritance trees in Search. one deals with "mechanism" of search.

etc

So if you see a declaration like

"SimulatedAnnealingSearch extends NodeExpander implements Search" , do not be confused.

==Logic Notes==

The ONE thing you need to watch out for is that the Parsers are VERY finicky . If you get a lexing or parsing error, there is a high probability there is an error in your logic string.

To use First Order Logic, first you need to create a subclass of aima.logic.fol.FOLDomain which collects the constants, predicates, functions etc that you use to solve a particular problem.

==LearningNotes==

===Main Classes and responsibilities===

A <DataSet> is a collection of <Example>s .Wherever you see "examples" in plural in the text , the code uses a DataSet . This makes it easy to aggregate operations that work on collections of examples in one place.

An Example is a collection of Attributes. Each example is a data point for Supervised Learning .