Bit-Selection Control for Energy-Efficient Handwritten Digits Recognition Hyperdimensional Computing Architecture

Abstract

Hyperdimensional computing (HDC) is a brain-inspired computing framework that provides simple and convenient methods to perform cognitive tasks like classification. Its foundation lies in the properties of very high dimensional vectors called hypervectors (HV). HDC is a promising alternative to the conventional von-Neumann architectures, but its high-dimensional processes still contain massive bit-wise operations. Current optimizations often sacrifice accuracy for better energy-efficiency. This work finds redundant bits in the associative memory that do not contribute any information during classification. A proposed bit-selection control trims these redundant bits leading to improved throughput and energy-savings without sacrificing accuracy. For the handwritten digits recognition problem, this simple control results in a 44.62% energy reduction at the cost of 8.34% increase in area.