A 65 nm Bayesian Neural Network Accelerator with 360 fJ/Sample In-Word GRNG for AI Uncertainty Estimation
This enables efficient uncertainty estimation for safety-critical edge applications like autonomous vehicles or medical diagnosis, though it is an incremental hardware optimization.
The paper tackles the high computational overhead of Bayesian neural networks (BNNs) for uncertainty estimation in AI by presenting an ASIC that integrates a 360 fJ/Sample Gaussian RNG into SRAM memory words, achieving 5.12 GSa/s RNG throughput and 102 GOp/s neural network throughput on a 0.45 mm2 chip.
Uncertainty estimation is an indispensable capability for AI-enabled, safety-critical applications, e.g. autonomous vehicles or medical diagnosis. Bayesian neural networks (BNNs) use Bayesian statistics to provide both classification predictions and uncertainty estimation, but they suffer from high computational overhead associated with random number generation and repeated sample iterations. Furthermore, BNNs are not immediately amenable to acceleration through compute-in-memory architectures due to the frequent memory writes necessary after each RNG operation. To address these challenges, we present an ASIC that integrates 360 fJ/Sample Gaussian RNG directly into the SRAM memory words. This integration reduces RNG overhead and enables fully-parallel compute-in-memory operations for BNNs. The prototype chip achieves 5.12 GSa/s RNG throughput and 102 GOp/s neural network throughput while occupying 0.45 mm2, bringing AI uncertainty estimation to edge computation.