4 Types of Soft Computing

John Spacey, May 14, 2018

Soft computing is an approach to software design that is tolerant of imprecision, uncertainty, partial truth and approximations. This is useful for problem spaces that are complex and/or that involve significant uncertainty. The following are common types of soft computing.

Fuzzy Logic

Fuzzy logic is a class of logic that allows for any probability as opposed to working with true and false.

Machine Learning

Software designed to learn by refining statistical models against data.

Heuristics

A class of algorithms based on fast approximations that aren't guaranteed to be precise or correct. For example, a search algorithm that uses a collection of ranking factors that can be computed quickly to guess at the best document for a search query.

Evolutionary Computation

A class of algorithms that solve problems with a process of trial and error. For example, a search algorithm that tries one heuristic but changes to another heuristic for a particular query if human's don't appear satisfied with the results.

Applications

Soft computing is useful a wide variety of applications:

- fast moving situations such as a self-driving car.

- complexities such as recognizing objects and people in images.

- situations with no "correct" solution such as an AI that needs to make a joke.

- areas that defy logic such as a natural language or emotion.

- flexibility such as a virtual customer service agent who can change its mind based on a customer argument.

Overview: Soft Computing
Type	Software Design
Definition	An approach to software design that is tolerant of imprecision, uncertainty, partial truth and approximations.
Related Concepts	Software Design » Uncertainty » Fuzzy Logic » Logic » Heuristics » Algorithms »