RRAM-Based CAM Combined With Time-Domain Circuits for Hyperdimensional Computing

Abstract

Content addressable memory (CAM) for search and match operations demands high speed and low power for near real-time decision-making across many critical domains. Resistive RAM-based in-memory computing has high potential in realizing an efficient static CAM for artificial intelligence tasks, especially on resource-constrained platforms. This paper presents an XNOR-based RRAM-CAM with a time-domain analog adder for efficient winning class computation. The CAM compares two operands, one voltage and the second one resistance, and outputs a voltage proportional to the similarity between the input query and the pre-stored patterns. Processing the summation of the output similarity voltages in the time-domain helps avoid voltage saturation, variation, and noise dominating the analog voltage-based computing. After that, to determine the winning class among the multiple classes, a digital realization is utilized to consider the class with the longest pulse width as the winning class. As a demonstrator, hyperdimensional computing for efficient MNIST classification is considered.The proposed design uses 65nm CMOS foundry technology and realistic data for RRAM with total area of 0.0077 mm2 , consumes 13.6 pJ of energy per 1k query within 10 ns clock cycle for 10 classes. It shows a reduction of ∼ 31× in area and ∼ 3× in energy consumption compared to fully digital ASIC implementation using 65nm foundry technology. The proposed design exhibits a remarkable reduction in area and energy compared to two of the state-of-the-art RRAM designs.

EventHD: Robust and efficient hyperdimensional learning with neuromorphic sensor

“…HDC has achieved comparable to higher accuracy compared to state-of-the-art machine learning models with lower execution energy. Much research also exploits the memory-centric nature of HDC to design in-memory acceleration platforms (Li et al, 2016 ; Halawani et al, 2021a , b ) However, existing HDC algorithms are often ineffective in encoding complex image data or keeping a notion of continuous-time. In contrast, we propose a novel method to preserve spatial-temporal correction, where spatial encoding keeps the correction of events in 2D space while temporal encoding defines correlation in a continuous-time dynamic.…”

GrapHD: Graph-Based Hyperdimensional Memorization for Brain-Like Cognitive Learning

“…However, operating over long binary vectors could still be costly or non-optimized for CPU and GPU platforms. CPUs do not have enough resources for parallelism, and GPUs are more suitable for highprecision computations such as floating-point values (Halawani et al, 2021;Imani et al, 2021;Poduval et al, 2021b). To accelerate GrapHD, we develop a novel platform that naturally operates over long binary vectors.…”