Figure 6

Numerical evidence of relevance of the quantum circuit in RL-RQAOA and separation between RL-RQAOA and RL-RONE. The above plot illustrates the separation between learning curves of RL-RQAOA and RL-RONE agents averaged across 15 bimodal weighted random 3-regular graphs with 100 (left) and 200 (right) nodes each. We chose \(n_{c}=10\) and \(n_{c}=18\) for 100 and 200 nodes, respectively in our simulations and the parameters \(\theta = (\alpha , \gamma , \vec{\beta})\) of the RL-RQAOA policy were initialized by setting \(\vec{\beta} = \{25\}^{{(n^{2}-n)}/2}\) and the angles \(\{\alpha , \gamma \}\) (at every iteration) to energy-optimal angles (i.e., by following one run of RQAOA). All agents were trained using REINFORCE (Alg. 1)