Mathematically Inclined
(I’ll postpone the Takao-san post a bit more, bear with me)
Last friday, I went with some people of my laboratory to visit a investment company. My current work is a optimization algorithm based on Memetic Algorithms applied to the Portfolio Optimization problem. When compared to other Evolutionary Computation-based approaches, it has a number of advantages (mail me if you are interested in the details).
However, talking to some people in the industry, the application side of my project seems based on a really old and outdated market model - not only mine, but most computer people doing portfolio optimization use the same market model as I do - the classical case of computer people not keeping in touch with the researchers of the field they are applying their algorithms to.
That should not be a big problem - just implement the new model, the optimization algorithm would work all the same, right? The problem is - according to this guy in the industry - they are pretty happy to use linear optimization methods to solve the new model. The new model is supposedly more complex than the old model I used. So either the old model was also solvable by linear models - something that most works before mine said was not possible - or the industry people haven’t been doing their mathematical homework correctly.
So, my current task is to prove that the new model either is not really solvable by linear methods (by examining its solution space, etc), or if I do find out that it is, then I have to shift the focus of my work to something else, like robot control methods or game playing - something that I wanted to do eventually for my thesis, but not just now.
Busy days loom ahead. :-/