| |
The term project on gene
mining is the most interesting part of the course. This time we
are asked to implement the mining algorithm on reversed linkage,
which is to find possible phenotypes caused by a given gene location
using the idea of association rules. Based on the basic idea of
downward and upward closure of association rules, we can design
quite different but efficient pruning policies. To date I've creatively
designed and implemented five different pruning policies, and
are working on the rest. Our current search space has been decreased
dramatically after the above pruning policies. In level four,
we decreased the size of candidate set from 6,000 to 2,000. And
in level five, we even decreased the size from 12,000 to less
than 500.
Although we designed and
implemented our own pruning policies, the initial ideas are from
the bunch of classic papers. You can find the list from our official
course webpage.
Later you can find project
report here.
|