Some days ago I asked Sids to create a separate board for Optimization Theory. I wanted to write about why I think we need this board among the other boards like distributed systems and IR which are seemingly more relevant to the research we do in OSL and Oktave. This is a generic meta post.
One thing is that most problems that we come across, if not all, are essentially optimization problems. It is a fairly obvious statement, but an important one. We need to find the most relevant content through search; we need to find the most efficient query plan; we need to design efficient operators that can analyse a huge body of webpages; we need to design better information systems; and so on. So, it is useful to know the different optimization techniques that are available.
The other thing important thing, at least to me, is that optimization theory begins from where our high school maths learning ended. (It should not have ended, but it did somehow, and rather abruptly.) It starts from the mathematical preliminaries like the following: limits, functions, continuity, single and multivariate calculus; linear algebra; vector calculus; geometry. It also develops the mathematical tools that make dealing with arbitrary number of dimensions simple. This is important because often the kind of data we deal with have huge number of dimensions. Also, the kinds of optimization problems that we try to solve involve multiple conflicting objectives and constraints.
Thirdly, it helps us convince ourselves that most real world problems cannot be solved! At least not in their pure forms. This is, to borrow a phrase from Sri, both unsettling and liberating at the same time.
Print