Advancements in Edge General Intelligence Architecture

Created using ChatSlide

This document explores Edge General Intelligence, emphasizing the need for local learning and adaptation in Edge AI. It presents a Dual-Timescale Neuromorphic Design featuring a 25 Hz planner and SNN control for continual adaptation. The architecture utilizes R-STDP synaptic pools and Knowledge Graphs for enhanced reasoning. Results indicate significant improvements in locomotion tasks and performance retention across domains, with gSCAN demonstrating compositional generalization at low power...