论文标题
二元神经网络的非局部优化
Nonlocal optimization of binary neural networks
论文作者
论文摘要
我们将训练二进制神经网络（BNN）作为因素图上的离散变量推理问题。我们研究了这种转换在参数较少的BNN设置中的行为，并提出了信仰传播的随机版本（BP）和调查传播（SP）信息传递算法以克服其当前配方的棘手性。与传统的BNN梯度方法相比，我们的结果表明随机BP和SP都可以找到BNN中参数的更好的配置。
We explore training Binary Neural Networks (BNNs) as a discrete variable inference problem over a factor graph. We study the behaviour of this conversion in an under-parameterized BNN setting and propose stochastic versions of Belief Propagation (BP) and Survey Propagation (SP) message passing algorithms to overcome the intractability of their current formulation. Compared to traditional gradient methods for BNNs, our results indicate that both stochastic BP and SP find better configurations of the parameters in the BNN.
