Make non-rewarding graph ends have a customisable reward #2
Description
Feature request.
Suppose when the agent reaches a non-rewarding graph end, instead of getting a reward of 0, you wanted to give it a small positive reward to see if it continues exploring to find the greater reward. Or, you could give it a small negative reward to encourage it to continue searching for the better rewarding end.
It would be nice to have the option available to put in the config json file, like we already have the high_reward_value and crash_reward_value.