Custom reward function


In the OpenAI Universe blog post ( they state that they extract rewards by “We developed a convolutional neural network-based OCR model that runs inside the Docker container’s Python controller, parses the score (from a screen buffer maintained via a VNC self-loop), and communicates it over the WebSocket channel to the agent.” Does anyone know if we can define a custom reward? The particular game that we are trying to run it on doesn’t have an on-screen score.