HEAM: High-Efficiency Approximate Multiplier optimization for Deep Neural Networks

We propose an optimization method for the automatic design of approximate multipliers, which minimizes the average error according to the operand distributions. Our multiplier achieves up to 50.24% higher accuracy than the best reproduced approximate multiplier in DNNs, with 15.76% smaller area, 25.05% less power consumption, and 3.50% shorter delay. Compared with an exact multiplier, our multiplier reduces the area, power consumption, and delay by 44.94%, 47.63%, and 16.78%, respectively, with negligible accuracy losses. The tested DNN accelerator modules with our multiplier obtain up to 18.70% smaller area and 9.99% less power consumption than the original modules.

💡 Research Summary

The paper introduces HEAM (High‑Efficiency Approximate Multiplier), a methodology for automatically designing approximate multipliers tailored to deep neural network (DNN) workloads. The authors observe that most existing approximate multipliers assume uniformly distributed operands, while quantized DNNs exhibit highly non‑uniform distributions for both weights and activations (often clustered around zero or mid‑range values). Ignoring these distributions leads to sub‑optimal error characteristics when the multipliers are deployed in neural network accelerators.

HEAM addresses this gap by explicitly modeling the joint probability distribution p(x, y) of the two operands and minimizing the expected squared error
E(θ) = ∑₍ₓ,ᵧ₎