Distributed Adaptive Softmax Policy Gradient Method with Variance Reduction: A Global Convergence Analysis | Synapse