Exact Neural Networks from Inexact Multipliers via Fibonacci Weight Encoding
TimeWednesday, December 8th11:30am - 11:50am PST
Event Type
Research Manuscript
Virtual Programs
Presented In-Person
Approximate Computing for AI/ML
DescriptionEdge devices must support computationally demanding algorithms, such as neural networks, within tight area/energy budgets. While approximate computing is demonstrated to alleviate these constraints, limiting induced errors remains an open challenge. In this paper, we propose a hardware/software co-design solution via an inexact multiplier, reducing area/power-delay-product requirements by 73/43%, respectively, while still computing exact results when one input is a Fibonacci encoded value. We then introduce a retraining strategy to quantize neural network weights to Fibonacci encoded values, ensuring exact computation during inference. We benchmark our strategy on Squeezenet 1.0, DenseNet-121, and ResNet-18, measuring accuracy degradations of only 0.4/1.1/1.7%, respectively.