User:Carrie55555/sandbox

Definition
Table driven agents use a percept sequence/ action table in memory to find the next action. They are implemented by a large lookup table.

A template for a Table Driven Agent: function TABLE-DRIVEN-AGENT(percept) returns action static: percepts, a sequence, initially empty static: table, a table, indexed by percept sequences, initially fully specified percepts <- APPEND(percept, percepts) action <- LOOKUP(percepts, table) return action

Specification
The case of “learning” is much different between humans and computer. To explain, Dix (2009) first look at what are the differences between human and computers. For humans, they have receiver organs such as vision, hearing and touching, these receivers give produce sense for human. Also, humans have reasoning and problem solving skills. Therefore, human can easily sense the environment nearby and use their problem solving skills to make suitable actions and learn from them. However, for computers, they do not have sensory organs and intelligent brains to analyse and solving the problem. They can only get data from some input devices such as mouse and keyboard and display the result by calculations or functions create by humans.

So, how do a computer or a machine “learn"? Intelligent agents were introduce for computers to define perceptions of dynamic conditions in the environment and determine suitable action (Hayes-Roth, 1994). One of the example of intelligent agent is table driven agent. The basic concept of this agent is to store every situation with the corresponding action in a table. When the computer meet a problem, it can just look for the solution for that situation stored in the action table.

Applications
One of the example using table driven agent is call management system in a company (Klingman, 2007). The system is used to manage calls (such as phone call) and link them to suitable replying agent. When the system receive a call, it checks the corresponding replying department in the company from the stored “redirecting table”, and find out whether the agent is available or not. If not, the call is being redirect to other department of the company, or even record the caller message in department mailbox automatically. The “redirecting table” is provided by the company themselves, and the company can change the content in the table whenever they want.

Limitations
As table driven agent uses a simple look-up table to determine the agent’s actions, it has to build a huge table to store the entries. For example, chess has about 10120 states. It takes a long time to build the table. Excessive size is one of the problem lead by a huge table, thus more storage memory is needed. This agent is inflexible too. It is not adaptive to changes in the environment (Klingman, 2007). When there is new changes, it requires entire table to be updated. It can’t make any actions conditional on previous actions or states. Therefore, table driven agent has no autonomy.

Reference

Dix, A. (2009). Human-computer interaction. In Encyclopedia of database systems (pp. 1327-1331). Springer, Boston, MA.

India University South Bend. Intelligent agents. Retrieved from http://www.cs.iusb.edu/~danav/teach/c463/3_agents.html

Klingman, E. E. (2007). U.S. Patent No. 7,187,662. Washington, DC: U.S. Patent and Trademark Office.

Romein, J. W., Plaat, A., Bal, H. E., & Schaeffer, J. (1999, July). Transposition table driven work scheduling in distributed search. In AAAI/IAAI (pp. 725-731).

San Diego Supercomputer Center(2018). Introduction to artificial intelligence. Retrieved from http://www.sdsc.edu/~tbailey/teaching/cse151/lectures/chap02.html