Soft Computing

Do we ever teach a child how to identify its parents? Well, probably we do to an extent. Now what would you say for the children of animals who have so less differences and still are able to make out each other out of the entire lot. Is it that the problem of differentiating people is so small. Probably not, since if it would have been so simple, we wouldn't have required the best of the scholarly community try whole-heartedly to make the machines solve a minute portion of it. Another comment that automatically comes in mind is that we are intelligent beings who are gifted intelligence by the All Mighty as the first Birthday Gift. In case that's true, why doesn't a child recognize things as vast as an adult does? A fact remains that we learn ourselves. This may be from our experiences, understanding, others experience, knowledge sharing practices, or any other means.

The innovation in intelligent systems hence does not necessarily lie in building systems that can carry out tasks, but to build a framework that enables them learn or what we say evolve along with changing needs, issues and complexities. The newly born has an abundant of things to learn and a relatively small time to do that. Imagine the learning as a result of constant reinforcement of vital concepts, short-term and long term learning, on-demand learning and the multiple forms that you may imagine. The systems are no different with they being asked to learn a bulk of important and non-important concepts for different times as per requirements. Hence we find a very strong co-relation with intelligent systems and intelligent humans that was largely due to the biological inspiration of computational intelligence.

Now when a student is asked to design systems to solve problems, he takes a very simple approach. Identify the procedures, methodologies, steps, algorithms, etc. A clear understanding of the problem and the adjoining methodology helps in effective solution that solves everything as far as the underlying assumptions are met. You ask the same student to make systems to solve any of the simplest natural problems and he might be buzzed. The big question is how would you be able to match two faces to be sure that they are actually the same person. Too many inputs, time of execution, tradeoffs, and in short complexity!!! We can solve the hardest of algorithmic problems but not even imagine to solve the simplest of natural problems.

The simple manner in which these are trivially solved motivate the need of alternate ways where any complexity can be cracked in the simplest of ways. This is a complete paradigm shift in the way we perceive problems, model them and ultimately solve them. Here the hardest of problems are solved as if they were a piece of pie cake. This is the world of Soft Computing. I do not claim the hard work of the people working in this field to be a showbiz (which also includes me), but claim the potential to make the impossible possible. Surely their work extracts the best out of the available limitations and restrictions, but the problem is ideally the problem is solved without any effort. The system or the framework does everything. So are the people fools who use complicated system modeling, algorithm design and other techniques to solve their problems? Well probably not. The systems we talk about here have a lot of assumptions regarding the large amount of data availability for experience and learning, continuous reinforcement, large computational access, etc. These may not always be the case with the other methodology of systems. Further the results are always approximate that are known to be correct rather than having an absolute correctness.

Now we know that the assumptions stated may not always hold good. This requires assistance in the various forms that creates a pool of possibilities of hybrid systems that are driven by multiple factors, each factor driving the system and enabling it suited for problem solving in its own way.

Myself on:

And also on:


Dr. Rahul Kala
Assistant Professor,
IIIT Allahabad,

Phone: +91 532 299 2117
Mobile: +91 7054 292 063