 |
Statistical
Inversion
Eigenor has a very fast
statistical inversion engine that can be used in a large variety
of problems. Statistical inversion methods typically give far better
results than traditional solving methods. And when there is only
a small amount of data, or when there is lot of noise, statistical
inversion is often the only possibility that can give acceptable
results. Furthermore, while using statistical inversion methods,
one can often have point-by-point estimates for the expected margin
of error in the result. This is something that the traditional methods
cannot provide.
The downside of statistical
inversion methods is the generally unacceptably large calculation
time. Eigenor’s statistical inversion engine can make many
of these slow calculations feasible to do, even in real time. In
an experiment few years back the same statistical inversion problem
for 3D reconstruction was solved at the Finnish supercomputing center
with their Cray supercomputer. Comparing it with Eigenor implementation
with then current state-of-the-art Mac computer, both calculations
gave the same result. However, the Cray with a commonly used implementation
needed 3.5 months to solve the problem, our solution did the same
in less than half an hour.
We have been doing digital
signal processing and analyzing radar signals with statistical inversion
in real time since 2001. Needless to say, this is a most remarkable
achievement.
Some of the application
areas for our statistical inversion engine include 3D reconstruction,
analyzing radar signals, analyzing geological survey data, analyzing
financial information, making
|
reconstructions
based on large amounts of measurement data from many complex
situations, or solving very large mathematical equations or
optimizations.
We like to think
that our difference is not only the greater speed or ability
to solve supercomputer class problems with fastest desktop
computers - or when supercomputers are available, ability
to solve even far greater problems. Although in scientific
research one may have the possibility to wait for several
months for results, this is not the case with industrial or
financial world applications. They need to be solved in real
time or in near real time. Otherwise there is likely no point
in doing them. In this respect, we believe our solutions can
open doors to a whole new world for our customer’s products
and solutions. |
|
