Difference between revisions of "Problem Solving with AI"

From GICL Wiki
Jump to: navigation, search
(Problem 3)
(Problem 4)
 
(5 intermediate revisions by one user not shown)
Line 6: Line 6:
 
Neural Networks are a standard for detecting credit-card fraud.  The neural network trains itself to identify normal spending patterns. They analyze each new purchase and compare it with previous purchases and raise a red flag when the purchases do not match the previous spending pattern.
 
Neural Networks are a standard for detecting credit-card fraud.  The neural network trains itself to identify normal spending patterns. They analyze each new purchase and compare it with previous purchases and raise a red flag when the purchases do not match the previous spending pattern.
  
To read more on Neural Networks please look here.
+
To read more on [[Artificial Neural Networks]]
  
 
== Problem 2 ==
 
== Problem 2 ==
Line 16: Line 16:
 
</p>
 
</p>
 
<p>
 
<p>
To read more on Fuzzy Logic please look here.
+
To read more on [[Fuzzy Logic]]
 
</p>
 
</p>
 
  
 
== Problem 3 ==
 
== Problem 3 ==
Line 26: Line 25:
  
 
''Autonomous Pedestrians'' is a study done by Wei Shao  and Demetri Terzopoulos at New York University. This study gives each individual pedestrian "Artificial Life" which controls thier movement, perception, behavior and thought to produce a life-like character (or agent). For more about Artifical Life [http://www.cs.ucla.edu/~dt/alife.html here]
 
''Autonomous Pedestrians'' is a study done by Wei Shao  and Demetri Terzopoulos at New York University. This study gives each individual pedestrian "Artificial Life" which controls thier movement, perception, behavior and thought to produce a life-like character (or agent). For more about Artifical Life [http://www.cs.ucla.edu/~dt/alife.html here]
 +
 +
==Problem 4==
 +
<p>As an animator you wish to have a character respond to physics and real world simutation while still reaching several constaints you put on it. This may range from balancing itself on the ground to responding to the weight of a caught object or reacting to getting hit by other characters or forces. What is the best AI to simulate real world responses without the need for hand animation?</p>
 +
<p>Evolutionary computation with its genetic algorithms is able to respond to contraints that change over time. A real-world application of this system is  [http://www.naturalmotion.com/ Natural Motion's Endorphin]
 +
To read more  on [[Evolutionary Computation]]</p>

Latest revision as of 10:47, 23 August 2007

The following are some problems that show the powers of diferent AI problem solving solutions

Contents

Problem 1

You are gathering data on the habits of a person, more specifically their spending habits. This data should help you build a profile on person who is spending so that any changes in their normal spending habits can easily be identified. To do this you should harness the power of computer intelligence. Which AI system works best?

Neural Networks are a standard for detecting credit-card fraud. The neural network trains itself to identify normal spending patterns. They analyze each new purchase and compare it with previous purchases and raise a red flag when the purchases do not match the previous spending pattern.

To read more on Artificial Neural Networks

Problem 2

You are creating a program that will act similar to a human but the information the program will get will not be exactly measurable (such as focusing a camera or keeping a fridge at a certain temperature) This program will need to take in approximate information and come to an answer that is best fit but not exact as such an answer would be impossible. What it the best AI system to use?

Fuzzy logic is utilized in many “automatic” and “smart” appliances today. Using imprecise data it makes approximations much the way humans do. Some common appliances that use fuzzy logic are refrigerators, ACs, washing machines and even car breaks.

To read more on Fuzzy Logic

Problem 3

You want to simulate a crowd of creatures interacting with architecture and eachother in a believable manner. An AI system will have to not only allow these crowds to avoid terrain and eachother, but also provide a crowd dynamic and a virtual goal to each character. What would be some possible soluations?

Massive is a software package created by Weta Digital that was used for the simualation of digital crowds it relies heavily on fuzzy logic. For more information on Massive you can look at thier website here

Autonomous Pedestrians is a study done by Wei Shao and Demetri Terzopoulos at New York University. This study gives each individual pedestrian "Artificial Life" which controls thier movement, perception, behavior and thought to produce a life-like character (or agent). For more about Artifical Life here

Problem 4

As an animator you wish to have a character respond to physics and real world simutation while still reaching several constaints you put on it. This may range from balancing itself on the ground to responding to the weight of a caught object or reacting to getting hit by other characters or forces. What is the best AI to simulate real world responses without the need for hand animation?

Evolutionary computation with its genetic algorithms is able to respond to contraints that change over time. A real-world application of this system is Natural Motion's Endorphin To read more on Evolutionary Computation