Guillemdb

from typing import Dict, Optional, Union

from fragile.core import DiscreteEnv
import gym
from gym import spaces
import numpy as np

from mathy import MathTypeKeysMax, MathyEnvState
from mathy.envs.gym import MathyGymEnv

Guillemdb

This game was played applying the causal entropic forces principles (http://www.alexwg.org/publications/PhysRevLett_110-168702.pdf).
We approximated the solutions of the proposed equation using the method described here (https://arxiv.org/abs/1705.08691). 

No neural networks were involved in this solutions. Everything was sampled using our "Fractal GAS" method, using only the cpu of
a laptop.  Since we did not implement any memory in the system, each game is independent of the other games played.

We would like some help, so we can use the games sampled using our method to train an A3C efficiently. As the "Fractal GAS"
allow us to sample human-like games from a uniform prior, we would be able to transform any RL problem into a supervised learning task. 

If you want us to share the code please show some interest and contact us in our blog http://entropicai.blogspot.com.es/