Formal Concept Analysis
For the last week or so I’ve been working on speeding up an algorithm for formal concept analysis (FCA) that I developed earlier in the year. FCA is a branch of lattice theory in which objects and properties are arranged into maximal sets of related entities. It’s like saying birds lay eggs, mammals nurse their young, but there’s a weird creature called the platypus that doesn’t really fit in either category.
We’re using FCA to build software plans from trace data collected on software that is run with multiple test cases. I developed the algorithm because a tool we were using (ToscanaJ) couldn’t handle the large input we were giving it. Since then we’ve learned of ConExp, which is supposed to be a lot faster. I did a quick comparison of my algorithm to their implementation of the NextClosure algorithm, and it looks like we might beat them. Meghan’s working on a more formal comparison.
Meanwhile I have to turn my attention to a paper… The ICSE formal demo paper deadline is fast approaching…