A key problem in nanomachine networks is how information from sensors is to be transmitted to a fusion center. In this paper, we propose a molecular communication-based event detection network. In particular, we develop a detection framework that can cope with scenarios where the molecules propagate according to anomalous diffusion instead of the conventional Brownian motion. We propose an algorithm for optimizing the network throughput by exploiting tools from reinforcement learning. Our algorithms are evaluated with the aid of numerical simulations, which demonstrate the tradeoffs between performance and complexity.