This work evaluates the performance of both Neural Networks (NNs) and Parameterized Quantum Circuits (PQCs) in Reinforcement Learning environments. We present novel quantum gate based games that...Show moreThis work evaluates the performance of both Neural Networks (NNs) and Parameterized Quantum Circuits (PQCs) in Reinforcement Learning environments. We present novel quantum gate based games that can run on current NISQ hardware. Our results show that NNs and PQCs achieve very similar performance distributions across the different environments, showing the promise of PQCs for (Quantum) Machine Learning applications. It is also shown that the NNs tested are more sensitive to change in learning rate than our PQC models. NN performance is also more eratic with relation to the amount of parameters than PQC performance, showing hyperparameter tuning might be more predictable for PQCs. Lastly, the smallest PQC designs show strong performance, often outperforming NNs with more parameters.Show less
