Home

DOCUMENTATION

The machine learning method used is non-deep neural network. Bayesian neural network is used to handle uncertainty of neural network’s weights (starting point is selected using L-BFGS 2nd order optimizer) so that uncertain cases can be filtered away from scoring results if wanted or targeted if risks are taken.

The 64bit C++ code is OpenMP multithreaded code using BLAS optimized linear algebra. This means software can use 100% of multicore CPU capacity. The software requires Windows 10 64bit and JRE 1.8 64bit Java. Other 64bit versions of Windows might work but are not tested.

KNOWN BUGS

1) non ASCII directory and filenames are not supported by C++ code.
2) Hybernation is not disabled by application so you need to disable it by hand.
3) About dialog tells this is commercial software but it is distributed now as freeware.
4) There are better ways to calculate scoring. Contact me if you willing to pay for better results.

For documentation with examples read doc from about menu.

Tomas.Ukkonen@NovelInsight.biz