Calculating the reward-rate manifold in the Gaussian-MSE case from Resource-rational reinforcement learning and sensorimotor causal states, and resource-rational maximiners

Calculates a reward-rate manifold when statistics are Gaussian and the reward function is a squared error. In this case, the generalized Blahut-Arimoto algorithm massively simplifies, and we can use the equations in Appendix D to find Fig 1.

当统计分布服从高斯分布且奖励函数为均方误差（squared error）时，可求解得到奖励率流形（reward-rate manifold）。在此场景下，广义布拉胡特-有马算法（generalized Blahut-Arimoto algorithm）的形式会大幅简化，我们可借助附录D中的公式推导得到图1。