In　this　paper,　a　novel　population-based　optimization　algorithm,　called　Free　Search　(FS),　is　studied.　First　the　essential　peculiarities　of　the　algorithm　is　introduced,　then　the　algorithm　is　improved　with　the　method　of　changing　search　neighbor　space　and　preserving　excellent　members　on　the　basis　of　sensitivity　of　the　algorithm　parameters,　thus　the　improved　Free　Search　Algorithm　(iFS)　is　proposed.　Some　canonical　equations　are　tested　with　experiments,　and　the　experimental　results　shows　iFS　can　speed　up　the　convergence　significantly　and　can　avoid　the　premature　convergence　effectively.　Compared　with　Free　Search　and　Genetic　Algorithm　(GA),　iFS　is　found　with　stable　robust　behavior　on　explored　results,　and　can　cope　with　heterogeneous　problems.　©　2009　IEEE.