By Eric M. Aldrich, Jesús Fernández-Villaverde, Ronald Gallant and Juan F. Rubio-Ramírez
http://d.repec.org/n?u=RePEc:pen:papers:10-014&r=dge
This paper shows how to build algorithms that use graphics processing units (GPUs) installed in most modern computers to solve dynamic equilibrium models in economics. In particular, we rely on the compute uni.ed device architecture (CUDA) of NVIDIA GPUs. We illustrate the power of the approach by solving a simple real business cycle model with value function iteration. We document improvements in speed of around 200 times and suggest that even further gains are likely.
There are fewer and fewer excuses for someone claiming a problem is too complex to solve. Computers have gained a lot of power, and this paper shows how to tap the unused part of your computer. Next up, the supposedly 90% of our brain we do not use.
Yeah, but try really global optimization, not local, or pretend it’s global with a super inadequate technique. That’s still going to often be very difficult with more realistic models, and is going to often rely more on intelligence than brute speed.