Integrating Event-Based Neuromorphic Processing and Hyperdimensional Computing with Tropical Algebra for Cognitive Ontology Networks

This paper presents a complete framework for combining event-based neuromorphic processing, hyperdimensional computing and tropical algebra for use within cognitive ontology networks. Using the Iris dataset I construct a virtual ontology network to simulate cognitive computing processes. Event-based neuromorphic processing models with spike activities and stochastic synapses dynamically adapt the networks’ topology. Hyperdimensional vectors represent the entities and relationships whilst tropical algebra operations bind these representations to encode complex relationships. A Multi-Layer Perceptron (MLP) with adaptive dropout and learning rates that are influenced by neuromorphic spike activities performs clustering and classification tasks.