BYNQNet: Bayesian Neural Network with Quadratic Activations for Sampling-Free Uncertainty Estimation on FPGA

High-Performance FPGA-based Accelerator for Bayesian Neural Networks

“…The high energy efficiency, computing capabilities and reconfigurability of FPGAs in particular make them a promising platform for acceleration of multiple different NN architectures [11]. Nevertheless, acceleration of BNNs specifically has not gained similar interests in the research community and there are only few works which approached this challenge [8]- [10].…”

“…However, their method needs to binarise the BNN and they again focus only on linear layers. Awano & Hashimoto [10] propose a custom inference algorithm for BNNs consisting exclusively of linear layers -BYNQNet which employs quadratic nonlinear activation functions and hence the uncertainty propagation can be achieved using only polynomial operations. Although the design can achieve a high throughput, the restriction of the nonlinear activation functions limits generality for different application scenarios.…”

“…However, because the current state-of-the-art NNs only spend a small portion of the execution time in the last layer, our accelerator can still achieve speed up in most of NNs, and thus is practical enough for real-life applications. We also compare our work with the other BNN accelerators [8], [10] in Table IV. Because the three-layer BNNs evaluated in [8] and [10] are unrealistic for real-life applications, we run a commonly-used ResNet-101 [12] on our design with MCD applied onto every layer, such that L = N .…”

“…We also compare our work with the other BNN accelerators [8], [10] in Table IV. Because the three-layer BNNs evaluated in [8] and [10] are unrealistic for real-life applications, we run a commonly-used ResNet-101 [12] on our design with MCD applied onto every layer, such that L = N . However, as both [8] and [10] do not support ResNet-101, their performance reported is still based on the three-layer BNN in their original papers.…”

“…necessary to approximate the predictive distribution through Monte Carlo sampling that requires the users to perform repeated sampling of random numbers and then run the same input data through the BNN multiple times, which degradates the hardware performance. Several algorithmic approximation techniques and hardware architectures [8]- [10] have been proposed to improve the hardware performance of BNNs. Nevertheless, there are particular drawbacks in these approaches: 1)…”