- Bahasa: Inggris
- Jumlah Halaman: 5
- Format file: PDF
- Publisher: Nagoya Institute of Technology, IlmuKomputer.Com
- Tahun terbit: January 2003
- Download thesis lengkap (PDF): anto-abstract-sc.zip
The
rapid progress in computer hardware industry has become a great
contribution to the progress of research in various fields. Many
computational problems with complex calculations can be solved within a
reasonable time, opening many inventions in a broad spectrum. Part of
these problems have clearly specified formulas and procedures to obtain
the solutions. The algorithms can be described very clear, from one to
the next steps, towards the final solution. In such cases, computer will
give very accurate result within very short computational time. It is
no doubt to conclude that in this sense, the performance of computer is
much superior to that of human brain.
A solution of a computational task, however, can
not always be stated clearly. Many problems in real-life domain are
always accompanied by imprecision or uncertainty factors. Sometimes the
information is far from complete, but a precise decision is required. In
such situation, human can make a correct decision, while computer
requires complex calculation to make a mathematical model of the
problem. One example of problems with such properties is handwriting
character recognition. According to the conventional methods described
in prior, we have to collect almost unlimited amount of representative
handwriting samples, and derive a certain formula or rule to make a
decision to which character that a handwriting pattern should be
classified. However, this kind of effort is not practical and the
formula will not tolerate to any deviation might occur in the
handwriting pattern of different persons. The character recognition
system should also tolerate the imprecision and uncertainty as common
characteristics of the handwriting characters. The solution to problems
with these properties is a family of algorithms named softcomputing.
