Search papers, labs, and topics across Lattice.
1
0
Recovering expert policies in mean-field games can be achieved with high fidelity using a novel maximum entropy approach, even in the absence of discount factors.