No abstract

Artificial intelligence tools have been heralded to pave the way to brave new powerful took for solving financial management problems. Often these would come in the form of expert systems or more recently in the form of artificial neural systems [ I ] . These tools however suffers from varous shortcomings. In the case of Expert Systems, extracting knowledge from experts have proved to be formidable tasks. In areas where igormation is inaccurate or incomplete, expert systems may be inadequatell] . Although Artificial neural systems ( U S ) offer the advantages of being able to cope with inaccurate or incomplete data and being able to learn from past data ( experience ), like a black box system it suffers from the inability to clearly explain the steps by which its decisions are reached. Also, although extracting all of the knowledge of the experts in building expert systems may be drffcult, some knowledge in the form of rules are fundamental and may be learnt from experts or from plain common sense. These fundamentals or rules of thumb must somehow be incoperated into these systems and ANS offers no graceful way for this to be accomplished. Many knowledge based systems such as expert systems are based on two value logic which forces a hard division in the decision space, ie., if something is not true than it must be false, nothing exist in between. This hard decision if is forced to be made early in a series of decisions may lead to cummulative errors and result in erroneous outputs. Fuzzy systems removes the necessity for this hard division. It resembles more closely the kind of common sense reasoning we use in decision making. In this paper , a stock selection strategy based on an artificial fwzy neural system (FNS) is describesd from a general system-design perspective. This paper suggests the concept of a neural gate which is similar to the processing element in ANS but generalized into handling various types of information such as fuzzy logic, probabilistic and Boolean iMormation combined. Forecasting of stock market returns, assessing of country risk and rating of stocks based on fuszy rules, probabilistic and Boolean data are areas where systems using theses neural gates may be applied. stock selection (ISS) system which extends the neural network approach to handle fuzzy, probabilistic and Boolean information. CH 3065-0/91/woo-1625 01.M) QIEEE

No abstract

No abstract

A novel Self-Tuning Adaptive Resolution (STAR) fuzzy control algorithm is introduced in this paper. One of the unique features is that the fuzzy linguistic concepts change constantly in response to the states of input signals. This is achieved by modifying the corresponding membership functions. We use this adaptive resolution capability to realize a control strategy that attempts to minimize both the rise time and the overshoot. Simulation results on a simple inverted pendulum problem are presented. Its characteristics are compared with the classical PD controller. Finally, the algorithm is also realized to control a red inverted pendulum hardware. Experimental results show that the STAR controller is both robust and can minimize positional error with drastically reduced overshoot. 0-7803-1896-X/94 $4.00 01994 IEEE 1508