|WikiProject Computer science||(Rated Start-class, Low-importance)|
|This article is the subject of an educational assignment at College Of Engineering Pune supported by Wikipedia Ambassadors through the India Education Program during the 2011 Q3 term. Further details are available on the course page.|
what is this soft computing? why we need this soft computing?
Perhaps the PAC learning framework should be listed as an example? I'm not sure. Alex.g 18:32, 8 December 2007 (UTC)
Adding the link probably approximately correct learning for reference. Actually, Zadeh does not mention this that I know of, so I don't think this would be the best example. I will be trying to clarify this page somewhat. --Clangin (talk) 13:58, 6 July 2008 (UTC)
Technically, SC is not AI
Artificial Intelligence (AI) uses hard computing techniques, whereas Soft Computing (SC) uses soft techniques. Therefore, they are not the same; they are in opposition. --Clangin (talk) 14:01, 6 July 2008 (UTC)
Artificial Intelligence uses both approaches (SC,HC)
It is very confusing for many people what IA is. A practical definition was proposed by Alan Turing, as a game where a person seating in a room by means of a teletype (that was in the early 50s) needs to guess if he is connected to either a computer or other human being in other teletype. If a person can not distinguish which one is the computer or the human, then the computer shows an artificial intelligence. In this accepted approach that I roughly described, there is nothing about the means by which such system may be achieved. Perceptrons (primitive neural networks), cellular automata, Turning machines, recursive functions, first order predicate logic, among others were known at that time. Also the difference between heuristic methods that always solve a class of problems, known as algorithms. Also methods that not always reach a solution or just approximates one, known as heuristic methods or simply as heuristics. The main difference in the way to find solutions resides in the representation of the problem. There were two approaches neaties and scruffies. Neaties use a neat representation of the problem, that means a symbolic representation of the problem, to which inference rules (rewriting) are used to find the solution. In that case what is needed to solve the problem is included as rules. That rules may be learned or programed. On the other hand, scruffies used methods inspired in the way the brain was supposed to work at that time. A simple method of a neuron was a perceptron, later that model evolved to neural networs, but the main difference was that there is no symbolic representation of the problem. In this approach a network of neurons is trained by presenting many samples of inputs that the system learns to classify. This approach is similar to the behaviorist school of psychology, were learning is a process that consist of presenting a stimulus and depending on the answer a positive or negative reinforcement, this is known as conditioning. Internally, the device works doing some kind statistics. The use of the expression "soft computing" seems to refer to the methods formerly known as scruffies, but I am not sure if they are synonymous, that was what leads me to this entry. If such is the case, a brief description of the subject may be placed in this article, and a link to bio-inspired methods, or something like that. By a previous analysis of the related terms. —Preceding unsigned comment added by Elias (talk • contribs) 12:00, 3 September 2010 (UTC)
- I don't think that definition for AI is clear and informative enough. I would propose a definition based on the characteristics that a system must display to show intelligence, one of them is possibly going to be the capability of (intelligently) adapting to circumstances, intelligence here means that the adaptation to circumstances or a different context is not programmed beforehand (for instance with a switch construct) but it is something obtained through the processing of the (intelligent) algorithm. Any system processes an input and produces an output. In the case of AI the input is either more unpredictable, more complex or both. The problem is that there is no consensus about a definition for AI.
- Back to the current topic, I'm not sure about the difference between neaties and scruffies, but certainly the difference between HC and SC is not the difference between symbolic and connectionist intelligence. Basically SC deals with approximations to solutions, either because a problem is undecidable or because the hard algorithm to find the solution (and not an approximation) is too expensive in computational terms. This difference between SC and HC is not very (co)related with the difference between AI and not AI, but if there is some correlation it is probably because the kind of problems where SC is interesting are those where AI is required (with a complex or unpredictable input that causes high computational costs or undecidability). --trylks (talk) 14:30, 14 March 2013 (UTC)
Needs a lot of work
I rewrote the introduction, which contained numerous factual errors, but this article really needs someone with a decent knowledge of the subject.
It needs a clear definition of what it is (I think my rewritten definition is reasonably accurate, but without any experience in the area I can't be sure), as well as a clear description of techniques used and its applications.
*goes to put "Expert needed" template on page*
hi I want to edit this soft computing page so can anyone provide me some more references to study this topic ? help me please. — Preceding unsigned comment added by Shrikrushna lekurwale (talk • contribs) 13:56, 30 September 2011 (UTC)