Partial Monitoring
A paradigm in sequential decision-making learning in which the learner selects an action each round but receives only a limited feedback signal. The learner can observe neither the true loss nor the actual environment outcome, and must decode the mapping between feedback symbols and true outcomes through repeated exploration.
Timeline (last 90 days)
The E2D.PMSide framework is the first to achieve computationally efficient solutions for contextual partial monitoring under general function approximation.
A partially monitoring learner makes a choice each round but receives only limited feedback about the outcome—it can neither observe the true loss nor know what outcome the environment actually selected.
The feedback structure of partial monitoring falls into three classic settings: globally observable, locally observable, and unobservable.
This work breaks the implicit assumption that 'partial monitoring can only use tabular methods', bringing general function approximation into the feasible toolbox for partial monitoring.
All Facts (4)
The E2D.PMSide framework is the first to achieve computationally efficient solutions for contextual partial monitoring under general function approximation.
50%UnverifiedA partially monitoring learner makes a choice each round but receives only limited feedback about the outcome—it can neither observe the true loss nor know what outcome the environment actually selected.
50%UnverifiedThe feedback structure of partial monitoring falls into three classic settings: globally observable, locally observable, and unobservable.
50%UnverifiedThis work breaks the implicit assumption that 'partial monitoring can only use tabular methods', bringing general function approximation into the feasible toolbox for partial monitoring.
50%