Mark Darguzas
John Mikulski
Intelligent Computing Techniques other than AI
Computing applications other than artificial intelligence (AI) are also making a major impact on business and information technology. Organizations are using computing techniques, such as neural networks, fuzzy logic, genetic algorithms, and intelligent agents, to expand their knowledge base. These computing techniques are able to work with problems and information which are too large or complicated for humans to handle, especially in a timely fashion. This knowledge management system will explain these techniques, the differences between them, and how they help organizations manage knowledge.
Instructions: Click on the links below to learn about a specific intelligent technique or aspect of it. Within each section, keywords are linked to their definitions in the glossary. Click on the keyword to read its definition.
Links:
* How the brain basically works
* How a neural network basically works
* Common applications of neural networks
* What is different about neural networks?
* More information on neural networks
* How fuzzy logic basically works
* Advantages of using fuzzy logic
* Common applications of fuzzy logic
* Example from Laudon and
Laudon
* Company Using
Hybrid AI Systems
* Distinguishing
factors for Agents
* Other
Application for Agents
As stated by Laudon & Laudon in Management Information Systems: Managing the Digital Firm, “neural networks – commonly called ‘neural nets’ – are designed to imitate the physical thought process of the biological brain.”
How the Brain basically Works: The biological brain is full of neurons, which are nerve cells that pass messages to each other by conducting impulses. The soma is at the center of the neuron and helps to stimulate other neurons. The “wire”, which carries the messages and information by connecting the neurons together, is called the axon. The axons are linked to dendrites, which are extensions of the neurons and transmit impulses inward toward the center of the cell. The point at which the axon and the dendrites meet is called the synapse. This is where the impulse passes from the axon to the nerve cell. This model is the basis for the development of neural networks.
How a Neural Network basically Works: Artificial neurons consist of processors called switches. Information is carried through wires, which act as the axons and dendrites. The information passes from the wire to the other processor (or artificial neuron) at the synapse, which is represented by variable resistors that carry weighted inputs (currents) that represent data.
The Neural Networks Training Problem
The
variable resistors in the circuits can be used to help the network learn. If the network makes a mistake by choosing
the wrong pathway through the network and arriving at the wrong answer, the
resistance can be raised on certain circuits.
This forces other neurons to fire, helping the network to eventually
come to the correct conclusion after this process is repeated for thousands of
cycles. Similar to the human brain, the
neural network can process massive amounts of data efficiently because the
neurons are highly interconnected and operate in parallel. A neural network uses rules it “learns” from
patterns in data to construct a hidden layer of logic.
Common Applications of Neural Networks: Applications of neural networks can be founding medicine, science, and business. These networks help address problems with pattern classification, prediction and financial analysis, and control and optimization.
Papnet is a neural net-based
system that helps technicians identify abnormal cells taken from Pap smears. The computer selects the abnormal cells and a
technician is able to review the computer’s selections. By using Papnet, a technician uses
approximately one-fifth the time to review a smear and is ten times as
accurate.
According to the Laudon & Laudon text, “Visa International Inc. is using a neural network to detect credit card fraud by checking all Visa transactions for sudden changes in the buying patterns of card holders.”
Neural networks are also being used in the financial industry. The networks are able to recognize patterns from giant databases that might help investment firms to predict the performance of equities, corporate bond ratings, or corporate bankruptcies. This information can allow financial investment firms to gain a strategic advantage.
What is Different about Neural Networks? Expert systems try to imitate a human expert’s way of solving problems; however neural network designers “try to put intelligence into the hardware in the form of a generalized capability to learn.” The expert system is made to solve a specific problem and is generally difficult to retrain. Another difference between expert systems and neural networks is that neural networks usually cannot explain how or why they came to a certain conclusion. Neural networks are also very sensitive to the amount of data they are trained with – too much or too little could greatly affect their level of performance. Finally, Laudon & Laudon states “neural networks are best used as aids to human decision makers instead of substitutes for them.”
More Information about Neural Networks:
Neural Networks: http://www.doc.ic.ac.uk/~nd/surprise_96/journal/vol4/cs11/report.html
Neural Networks At Your Fingertips: http://www.neural-networks-at-your-fingertips.com/
According to Laudon & Laudon, “fuzzy logic is a rule-based technology that tolerated imprecision and even uses it to solve problems we could have not solved before.”
How Fuzzy Logic basically Works: Fuzzy logic is able to create rules by inferring knowledge from imprecise, uncertain, or unreliable information. Programmers use imprecisely defined terms, which are known as membership functions. These membership functions are a series of IF-THEN rules; however, fuzzy logic code requires fewer IF-THEN rules than traditional code, which makes it simpler to use and to write. The computer asks the user all questions, then combines the membership function readings in a weighted manner, and finally makes a decision based on the user’s answers to all questions.
http://www.fuzzy-logic.com/Ch1.htm
An Example of Fuzzy Logic: A computer is connected to the thermostat in a room, allowing it to control the room’s temperature and level of humidity. In order for the computer to make a decision, rules based on imprecise definitions for indoor temperature (cool is between 50 and 70 degrees, yet cold can be below 60 degrees), humidity, outdoor wind, and outdoor temperature. One rule could be as follows: “If the temperature is cool or cold and the humidity is low while the outdoor wind is high and the outdoor temperature is low, raise the heat and humidity in the room.”
Advantages of Using Fuzzy Logic:
Managers who use fuzzy logic have found that it can reduce costs and
shorten development time. Again, fuzzy
logic code requires fewer IF-THEN rules than traditional code. This compact code, therefore, uses less
computer storage space. Fuzzy logic also
allows humans to solve problems not previously solvable, which helps to improve
product quality. This computing
technique can also be useful for making better decisions and better controlling
the organization.
Common Applications of Fuzzy Logic:
Fuzzy logic is more commonly used in
* Sanyo Fisher USA uses fuzzy logic to implement controls and autofocus devices in camcorders.
*
In
*
Mitsubishi Heavy Industries in
*
Williams-Sonoma sells an “intelligent” steamer
made in
* A Wall Street firm had a system developed that selects companies for potential acquisition, using the language stock traders understand.
*
A system has been developed to detect possible
fraud in medical claims submitted by healthcare providers anywhere in the
What is Genetic Algorithm: the program is designed for problem solving based off the evolution process. The program continually re-adjusts, reorganizes and even mutates to continually find a better solution. What interesting about Genetic Algorithms is the program is essentially breeding new designs.
Example from Laudon and Laudon (page 337): General Electric uses genetic algorithms to design the best jet turbine aircraft engine. This process involves changing and reorganizing component parts of the engine. This is done in computer simulator. After parts were change or reorganized, the computer would measure which new design yielded the best engine. Then, the new design repeats this evolution process again. This continually happens until the computers designs don’t yield better results. The new breeds of are then looked at by the engineer. The designs are developed much faster than if a team were to sit down and test our each design.
What are Hybrid AI Systems: According to Laudon & Laudon, hybrid AI systems are combined genetic algorithms, fuzzy logic, neural networks and / or expert systems to further optimize the programs problem solving capabilities. Number of companies combining these software are expanding.
Company Using Hybrid AI Systems: Matsushita developed a washing machine that combines neuro-networks with fuzzy logic. The washing machine is user friendly and because of the software, achieves a new level of energy efficiency and shortens drying time.
What are Intelligent Agents: are applications that perform repetitive tasks, without being managed by a human. The agents are also referred to as “robots”, “bots”, “crawlers” and “worms”. A popular use for the software is shopping. Shopping bots automatically check web-sites for the best prices. Then alert the user through email about the best price. Also, another popular use for agents are from websites such as Monster. At Monster, the bots automatically check the job posts weekly. After the bots or agent finds a post that matches a description from the job seeker, the program alerts the user through email. This prevents the job seeker from check continuously. The agent saves the use a large amount of time.
Distinguishing factors for agents: These factors mostly have to do with intelligent and not intelligent agents. How well the agent can adapt to fit the users needs. Also, how well the agent can adapt based off its performance. If search results are not yielding pertinent information, the searching software has to choose another strategy. Moreover, how pro active the agent is. For example email the user after the agent found a target. Additionally, how user-friendly the agent is. Lastly, distinguished factor is function the agent is performing.
Example of Shopping bots: The bots on this page search out product and price information from online stores and then report back, by email, to the user. The agents can adapt. The focus can be on customer reviews to shipping needs. By delegating the agent to check prices for example, the user can spend time on school, becoming more efficient. (http://dir.yahoo.com/Business_and_Economy/Shopping_and_Services/Retailers/Virtual_Malls/Shopping_Agents/).
Other Application for Agents: 1, email filter 2, appointments scheduler 3, cheap air-fair finder
4, stock screeners 5, job finders 6, ect.
Artificial intelligence (AI) – the
effort to develop computer-based systems that can behave like humans, with the
ability to learn languages, accomplish physical tasks, use a perceptual
apparatus, and emulate human expertise and decision making
Axon – “wire”, which acts as an electrically active link to the dendrites of other neurons
Dendrite – a branched protoplasmic extension of a nerve cell that
conducts impulses from adjacent cells inward toward the cell body. A single
nerve may possess many dendrites. Also called dendron (http://dictionary.reference.com/search?q=dendrite)
Expert system –
knowledge-intensive computer program that captures the expertise of a human in
limited domains of knowledge
Fuzzy logic – rule-based artificial intelligence (AI) that tolerates imprecision by using nonspecific terms called membership functions to solve problems
Genetic Algorithms -Problem-solving methods that promote the evolution of solutions to specified problems using the model of living organisms adapting to their environment.
Hybrid AI Systems - Integration of multiple AI technologies into a singe application to take advantage of the best features of these technologies.
Intelligent Agent - Software program that uses a built-in or learned knowledge base to carry out specific, repetitive, and predictable task for an individual user, business process, or software application.
Membership function
– the imprecisely defined terms used in fuzzy logic
Neural network – hardware or software that attempts to emulate the processing patterns of the biological brain
Neuron – any of the impulse-conducting cells that constitute the
brain, spinal column, and nerves, consisting of a nucleated cell body with one
or more dendrites and a single axon. Also called nerve cell (http://dictionary.reference.com/search?q=neuron)
Papnet – a neural net-based system that
distinguishes between normal and abnormal cells when examining Pap smears for
cervical cancer
Soma – nerve cell at the center of a neuron that acts like a switch, stimulating other neurons and being stimulated in turn
Synapse – the junction across which a nerve impulse passes from an axon terminal to a neuron, muscle cell, or gland cell (http://dictionary.reference.com/search?q=synapse)