Free Trial

Safari Books Online is a digital library providing on-demand subscription access to thousands of learning resources.


Share this Page URL
Help

INTRODUCTION > INTRODUCTION - Pg. 219

Fuzzy Logic in Medicine series of propositions into mathematical proofs. Computers are now able to solve proofs in ways never previously conceived. Fuzzy logic is similar to classical logic in the search for truthfulness of a proposition. Some- times truth is subjective ill defined. As an example, it is difficult to assign a true or false value to the proposition "Andy is tall" or "Shell is old." How tall does one have to be before being categorized as "tall"? Likewise, how old does one have to be before being considered "old"? Most would agree that 100 years is old for a person but young for a planet. Like many real-world propositions, the concept of age is relative to its usage. To solve these problems there was a need to develop a more robust system of logic. Rather than assigning a proposition as either 0 or 1 the idea of variable truth was added. The variable is measured over the interval of [0, 1]. Fuzzy logic rose from this concept. One major focus of this discipline is in Many techniques have been used to create fuzzy logic programs that function as an expert. The earliest systems used conditional statements with tolerance thresholds using if-then-else rules (Jackson, 1999). This approach, while seemingly simplistic, has been used successfully in a wide va- riety of medical applications including diagnostics and psychological bias (Shortliffe, 1976). Other less known approaches of fuzzy logic systems are association nets and frames, which have proven difficult to implement with only marginal results. The two most common implementations of fuzzy logic are rule-based and neural networks. Both fuzzy implementations have a diverse range of ap- plications including medicine, avionics, security, and machine learning. Unlike rule-based fuzzy logic, neural nets do not require thinking patterns to be explicitly specified. Typically two data sets are created to program a neural network. The first data set is