yasufumy/python_irl

Basic Inverse Reinforcement Learning method in Python

python_irl's Issues

__call__ in Reward class

I am afraid that there is a bug in call function in class Reward, maxent_lrl.py.
Dealing with feature_matrix when it is int, the present expression is:
feature_matrix = np.eye(self.n_features, 1, feature_matrix).T
This returns [1, 0, 0, ..., 0] (all zero except the first item is 1), or [0,0,...,0] (all items are zero).
I think that line should be like this, for example,
feature_matrix = np.eye(self.n_feature)[feature_matrix)
Thank you for your consideration.