Brains perform timely reliable decision-making by Bayes theorem. Bayes theorem quantifies events as probabilities and, through probability rules, renders the decisions. Learning from this, applying Bayes theorem in practical problems can visualize the potential risks and decision confidence, thereby enabling efficient user-scene interactions. However, given the probabilistic nature, implementing Bayes theorem with the conventional deterministic computing can inevitably induce excessive computational cost and decision latency. Herein, we propose a probabilistic computing approach using memristors to implement Bayes theorem. We integrate volatile memristors with Boolean logics and, by exploiting the volatile stochastic switching of the memristors, realize Boolean operations with statistical probabilities and correlations, key for enabling Bayes theorem. To practically demonstrate the effectiveness of our memristor-enabled Bayes theorem approach in user-scene interactions, we design lightweight Bayesian inference and fusion operators using our probabilistic logics and apply the operators in road scene parsing for self-driving, including route planning and obstacle detection. The results show that our operators can achieve reliable decisions at a rate over 2,500 frames per second, outperforming human decision-making and the existing driving assistance systems.

翻译：大脑通过贝叶斯定理实现实时可靠的决策。贝叶斯定理将事件量化为概率，并通过概率规则进行决策。受此启发，在实际问题中应用贝叶斯定理可以可视化潜在风险和决策置信度，从而实现高效的用户-场景交互。然而，鉴于其概率特性，使用传统的确定性计算实现贝叶斯定理不可避免地会导致过高的计算成本和决策延迟。本文提出了一种基于忆阻器的概率计算方法来实现贝叶斯定理。我们将易失性忆阻器与布尔逻辑相结合，通过利用忆阻器的易失性随机切换特性，实现了具有统计概率和相关性的布尔运算——这是实现贝叶斯定理的关键。为实际验证我们基于忆阻器的贝叶斯定理方法在用户-场景交互中的有效性，我们使用概率逻辑设计了轻量级贝叶斯推理与融合算子，并将这些算子应用于自动驾驶道路场景解析，包括路径规划和障碍物检测。结果表明，我们的算子能以超过每秒2,500帧的速率实现可靠决策，其性能优于人类决策及现有驾驶辅助系统。