Identifying the Reward Function using Anchor Actions