Numerical Simulation

Cross-source consensus on Numerical Simulation from 1 sources and 6 claims.

1 sources · 6 claims

Benefits

The simulations use eta = 1/275, which lies in the global-clock unstable interval. — On the Divergence of Differential Temporal Difference Learning without Local Clocks
The simulation rewards are always zero. — On the Divergence of Differential Temporal Difference Learning without Local Clocks
The reported simulations instantiate the smallest counterexample with m = 23 and 25 states. — On the Divergence of Differential Temporal Difference Learning without Local Clocks
The local-clock recursion contracts toward zero in the simulation. — On the Divergence of Differential Temporal Difference Learning without Local Clocks
The global-clock recursion grows and oscillates in the simulation. — On the Divergence of Differential Temporal Difference Learning without Local Clocks
Additional random seeds showed the same qualitative simulation behavior. — On the Divergence of Differential Temporal Difference Learning without Local Clocks