NEAR-SADDLE-NODE BIFURCATION BEHAVIOR AS DYNAMICS IN WORKING-MEMORY FOR GOAL-DIRECTED BEHAVIOR

In consideration of working memory as a means for goal-directed behavi or in nonstationary environments, we argue that the dynamics of workin g memory should satisfy two opposing demands: long-term maintenance an d quick transition. These two characteristics are contradictory within the linear domain. We propose the near-saddle-node bifurcation behavi or of a sigmoidal unit with a self-connection as a candidate of the dy namical mechanism that satisfies both of these demands. It is shown in evolutionary programming experiments that the near-saddle-node bifurc ation behavior can be found in recurrent networks optimized for a task that requires efficient use of working memory. The result suggests th at the near-saddle-node bifurcation behavior may be a functional neces sity for survival in nonstationary environments.