Sparsely Connected Associative Memory with Adaptive Topology through Annealed Dilution

A novel sparsely connected associative memory model with adaptive topology through annealed dilution is proposed in this paper. Aimed at overcoming the disadvantage of quenched dilution as random synapses disconnection of the existing methods, this new model takes the ideology of annealed dilution into account, and investigates the optimal synaptic dilution strategy under the constraints of limited amount of neurons and connections. Based on explicit theoretical analysis, this new model breaks the traditional manner of quenched dilution but instead constructs a learning task-dependent network topology through annealed way which is much closer to biological genuine system as possessing flexible adaptive structure and can also achieve better performance than the existing counterparts of the same class. the rationality and validity of the proposed model is validated from great number of experiments.