CS 453: Artificial Intelligence, Spring 2005
Information About the Third Test

The third and final test for this course will be given in class on Friday, April 29. It will cover the following material:

There will be no programming on this test. You can expect some questions on neural nets, but you do not need to know all the mathematical detail from Section 20.5. There will probably be something on knowledge representation using specific techniques such as sematic nets, predicate logic, or Lisp expressions. There will be no detailed questions about formal grammars and parsing, but you should be familiar with the general idea, including semantic grammars. Other than that, the test will consist of definitions, short-answer questions, and essay questions.

Here are some of the terms and ideas that you should be familiar with:

   Predicate logic                            natural language processing                           
   constant symbols                           natural vs. artificial languages                      
   function symbols                           problem of understanding                              
   predicates                                 machine translation                                   
   logical operators                          natural language interfaces                           
   for-all and there-exists                   acoustic level                                        
   variables                                  phonemes                                              
   models for predicate logic                 morphemes                                             
   entailment in predicate logic              lexemes (words)                                       
   expressing "isa" in logic                  syntax                                                
                                              semantics                                             
   knowledge representation                   pragmatics                                            
   production rules                           grammars and parsing                                  
   concept hierarchies                        ambiguity                                             
     "isa"                                    "times flies like an arrow"                           
     "instance of"                            semantic grammars                                     
     "part of"                                semantic primitives                                   
     inheritance in isa hierarchies              (PTRANS, ATRANS, MTRANS...)                        
   frames                                     using frames to represent semantics                   
   slots and fillers                          using semantic nets for semantics                     
   default fillers                            scripts                                               
   semantic nets                              difficult issues in semantics                         
                                                 quantity                                           
   learning                                      time and space                                     
   learning versus memorizing facts              knowledge and belief                               
   types of learning                             pronoun references                                 
      supervised (training examples)             conversational context                             
      unsupervised (concept discovery)                                                               
      reinforcement (reward/punishment)       the Turing Test (aka "imitation game")  
   learning classifications                   arguments against AI                                  
   learning concepts                             the mathematical objection                         
   Ockham's Razor                                the argument from consciousness                    
   decision tree                                 arguments from various disabilities                
   IDE3 algorithm                                Lady Lovelace's objection                          
                                                 argument from continuity in the nervous system     
   neural nets                                   argument from informality of behavior              
   the basic neural net node                                                                        
   inputs and output
   weights
   bias input
   activation function
   threshold functions
   sigmoid function
   McCulloch-Pitts neuron
   perceptron
   the perceptron learning rule
   learning rate (alpha)
   hidden layers
   back-propagation learning