post_parent): $temp_content = $post->post_content; $temp_content = explode("

",$temp_content); $temp_content = $temp_content[1]; $temp_content = explode("

",$temp_content); $temp_content = $temp_content[0]; $temp_content = strip_tags($temp_content); $temp_content = trim($temp_content); $authors = explode(",",$temp_content); ?> post_title));?>"> $value): ?> "> "> "> post_content); while ($parser->parse()) { if (($parser->iNodeName=="a")&&(substr_count($parser->iNodeAttributes['href'],".pdf")>0)): ?>

On performing actuarial calculations on GPU

Norkin B.V.

The paper considers issues of performing actuarial calculations on graphical processing units (GPU). Ruin probability estimation for an insurance company can always be made by means of Monte-Carlo method. In some cases it is the only applicable method. Smallness of ruin probability may require astronomical number of Monte-Carlo simulations to achieve an acceptable accuracy. Effective parallelization of the Monte Carlo method and utilizing calculation power of modern GPU makes it possible to obtain good accuracy within reasonable real time limits. A software system RMS 0.1 utilizing NVIDIA CUDA 4.2 architecture for modeling reserve evolution and performance of an insurance company was developed. Results of numerical experiments by means of this system are presented.


Full text (pdf)