TreeGRNG: Binary Tree Gaussian Random Number Generator for Efficient Probabilistic AI Hardware
Researchers have developed TreeGRNG, a novel binary tree Gaussian random number generator designed for efficient probabilistic AI hardware. This innovation addresses the significant power and computational demands of traditional Gaussian random number generators used in Bayesian Neural Networks. TreeGRNG achieves a 3.7x reduction in energy per sample and a 5.8x increase in throughput per unit area, while also offering greater flexibility in adjusting probability distributions. AI
IMPACT This development could lead to more efficient and trustworthy AI hardware, particularly for edge devices running probabilistic models.