New paper published

A new paper of my group was published today on Lecture Notes on Computer Science. It is a conference paper that we presented at the 2017 International Meeting on Computational Intelligence Methods for Bioinformatics and Biostatistics.

The paper, titled “Estimation of Kinetic Reaction Constants: Exploiting Reboot Strategies to Improve PSO’s Performance”, investigates the impact of reboot strategies on the estimation of kinetic parameters of biochemical reaction networks by means of Particle Swarm Optimization.

We investigated this possibility and we concluded that: yes, reboots are very helpful and improve the performances; the “local” strategy (i.e., a particle not improving its performances for too many iterations is randomly relocated in the search space) is the most performant algorithm.

Even though rebooting seems to be effective, it is actually not exactly trivial to apply to self-tuning algorithms like FST-PSO (as it confuses the fuzzy reasoning) and we are still investigating this idea.