A Dynamic Hebbian Learning Algorithm for Constructing E-Learner Communities

In order to solve the problem of "information overflow" in e-learning, an algorithm based on Hebbian learning law is proposed for constructing self-organized communities which can automatically group e-learners according to their learning interests. Unlike filtering methods,this algorithm takes into consideration of the distributed open environment of e-learning. This paper designed a peer-to-peer architecture and applied Hebbian learning law in constructing e-learner communities, avoiding the difficulty in calculating user similarity. Compared with traditional Hebbian learning based algorithm, this algorithm uses dynamic thresholds to address the problem of unilateral trust weight adjustment in extreme cases, and it also improves the trust weight adaptation and neighbor adjustment policies. Experimental results show that this algorithm achieves faster community construction speed and better scalability and stability.