re-in-pid

To use it within your reinforcement learning context, you need:

• https://github.com/Abzinger/BROJA_2PID (BROJA)
• import dit (Finn-Lizier)

After that, use the function

• intrinsic_reward(n, piT, piX_T, piY_t)

which takes as arguments

• n: The number of actions
• piT: some array-ish object such that piT[t] is the probability that T takes action t
• piX_T: such that piX_T[x,t] is the probability, conditioned on T taking action t, that X takes action x
• piY_T: such that piY_T[y,t] is the probability, conditioned on T taking action t, that Y takes action y

The function returns a single floating point number normalized to [-1,+1].

To install this package just do:

pip install git+https://github.com/dojt/re-in-pid