A reinforcement learning approach to online clustering

A general technique is proposed for embedding online clustering algorithms based on competitive learning in a reinforcement learning framework. The ba sic idea is that the clustering system can be viewed as a reinforcement lea rning system that learns through reinforcements to follow the clustering st rategy we wish to implement. In this sense, the reinforcement guided compet itive learning (RC;CL) algorithm is proposed that constitutes a reinforceme nt-based adaptation of learning vector quantization (LVQ) with enhanced clu stering capabilities. In addition, we suggest extensions of RGCL and LVQ th at are characterized by the property of sustained exploration and significa ntly improve the performance of those algorithms, as indicated by experimen tal tests on well-known data sets.