A STDP-based Encoding/Decoding Algorithm for Associative and Composite Data

Hong-Gyu Yoon, Pilwon Kim

Spike-timing-dependent plasticity(STDP) is a biological process of synaptic modification caused by the difference of firing order and timing between neurons. In our separate work, we have rigorously shown that STDP in a network of neurons transforms periodic input patterns into a geometrical structure named memory plane, and a proper memory cue near such structure dynamically revives the stored information. Using these results, in this paper, we demonstrate the following two encoding/decoding algorithms handling practical data. First, we perform an auto-associative memory task with a group of images. The results show that any relevant image to the group can be used as a cue in order to reconstruct the original images. The next one deals with the process of semantic memory representations that are embedded from sentences. The results show that words can recall multiple sentences simultaneously or one exclusively, depending on their grammatical relations. This implies that the proposed framework is apt to process multiple groups of associative memories with a composite structure.

Knowledge Graph

arrow_drop_up

Comments

Sign up or login to leave a comment