Curiosity

class Curiosity(icm: pandemonium.implementations.icm.ICM)

Bases: pandemonium.cumulants.Cumulant

Measures the novelty using prediction error of the forward model.

Intrinsic reward on the transition at time \(t\) is given by

\[r^i_t = \frac{1}{2} \norm{\hat{\phi}(s_{t+1}) - \phi(s_{t+1})}^2_2\]

Methods Summary

__call__(self, experience, ForwardRef])

Call self as a function.

Methods Documentation

__call__(self, experience: Union[ForwardRef(‘Transition’), ForwardRef(‘Trajectory’)])

Call self as a function.