This project proposes a novel error correction method based on deep reinforcement learning. Errors in quantum annealers manifest themselves as a suboptimal solution to a given optimisation problem.  However, even the wrong answer is relatively ‘close’ to the correct one.  The main idea of this project is to present the error correction process as a single-player game in which the player tries to ‘improve’ a given solution. In this way, the error correction task can be formulated as a reinforcement learning problem, where the player learns by trial and error to identify and correct the erroneous parts of the solution. Due to the enormous success of AlphaZero and derived algorithms (e.g. AlphaDev), it is planned to base the described method on this algorithm. The proposed game is a combinatorial game (in the sense of game theory), for which AlphaZero is very well suited. It is also planned to utilise the graph structure of a quantum processor by using graph neural networks.