Optimization using multiple processors

Not planned

Comments

2 comments

  • Avatar
    Tony Zheng

    I was wondering the same. Some of the optimization process can take far too long. Also is there anything better than Box? (There must be as Box's method is from 1960s)

  • Avatar
    Rick Kossik

    Two different questions here:  1) using multiple processors for optimization: yes, this is something that could indeed be programmed. Unfortunately, it is by no means a trivial design and programming effort.  2) Is there something better than Box's method? Certainly.  But Box's method is simple, straightforward, well-known, and generally pretty good.  Of course, there are more advanced methods (e.g., genetic algorithms) that could likely find global optima better.  In fact, there are companies who have spent a lot of time developing these and sell them for a living.  Hence, if we did it, we would not reinvent the wheel (it would be foolish to do so). In fact, we looked into this about 10 years ago. The issue in that case was that it would have added at least $500 to $1000 to the price tag for GoldSim (that was the licensing fee at that time). Would people pay that?  Bottom line is that both of these requests are good ones (and have been on our "to-do" list for a long time), but both are non-trivial (and in the case of new optimization methods, would make GoldSim more expensive), and hence are good examples of how (given limited resources), when we add new features, we strive to select features that reach a large number of users (rather than just a small number).  In general, advanced optimization is just not something lots of people have clamored for.  However, if that were to change, it then may be worthwhile to pursue it.

Please sign in to leave a comment.