v6.3 [Aug 16, 2014]
Random Numbers & Sampling
- Added a pseudorandom number generator, extensively tested with the Statistical Test Suite from the National Institute of Standards – great for financial, stochastic or statistical models that are heavy consumers of random numbers.
- Optimization and sensitivity control files may now specify a random seed of 0 to trigger randomization to system cryptographic sources (useful for performing independent runs on multiple computers without editing the control files).
- Improved sampling for Sensitivity (Monte Carlo) Simulations.
Optimization
- Big improvements to Markov Chain Monte Carlo and Simulated Annealing performance
- A broader distribution of initial test points in random multistart optimization (MWYS), to find optimal solutions faster (up to 100x in one test case).
- Improved speed in stochastic optimization, with support for the File sensitivity control option.
