MajorityNets: BNNs Utilising Approximate Popcount for Improved Efficiency
This work addresses resource constraints in embedded implementations of BNNs, offering an incremental improvement for FPGA-based systems.
The paper tackles the inefficiency of XnorPopcount operations in binarized neural networks (BNNs) by proposing XNorMaj, an approximate replacement that is up to 2x more resource-efficient, enabling the use of larger networks to recover minor accuracy losses.
Binarized neural networks (BNNs) have shown exciting potential for utilising neural networks in embedded implementations where area, energy and latency constraints are paramount. With BNNs, multiply-accumulate (MAC) operations can be simplified to XnorPopcount operations, leading to massive reductions in both memory and computation resources. Furthermore, multiple efficient implementations of BNNs have been reported on field-programmable gate array (FPGA) implementations. This paper proposes a smaller, faster, more energy-efficient approximate replacement for the XnorPopcountoperation, called XNorMaj, inspired by state-of-the-art FPGAlook-up table schemes which benefit FPGA implementations. Weshow that XNorMaj is up to 2x more resource-efficient than the XnorPopcount operation. While the XNorMaj operation has a minor detrimental impact on accuracy, the resource savings enable us to use larger networks to recover the loss.