Alpha-divergence minimization with mixed variational posterior for Bayesian neural networks and its robustness against adversarial examples
Neurocomputing(2021)
摘要
•We develop alpha-divergence minimization with mixed variational posterior.•The proposed method is time-efficient and extendable to various neural networks.•We provide the uncertainty comparison of Bayesian neural networks adopting different posterior approximations.•Experimental results show that our method achieves higher performance.
更多查看译文
关键词
Bayesian neural networks,Alpha divergence,Approximate inference
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要