Anytime Decision Metropolis-Hastings sampler.
Trait that defines some common interface functions for decision algorithms.
Importance sampling for decisions.
Importance sampling for decisions. Almost the exact same as normal importance sampling except that it keeps track of utilities and probabilities (to compute expected utility) and it implements DecisionAlgorithm trait.
Metropolis-Hastings Decision sampler.
Metropolis-Hastings Decision sampler. Almost the exact same as normal MH except that it keeps track of utilities and probabilities (to compute expected utility) and it implements DecisionAlgorithm trait
Abstract base class for all Decision Policies.
Abstract base class for all Decision Policies. Must define two functions: toFcn: T => Element[U] - this is the function that is called to compute the decision for a parent value.
toUtility: T => Element[Double] - this returns the expected utility of the decision for a parent value. Used in backward induction algorithm.
The parent value type
The decision type
An exact decision policy.
An exact decision policy. This policy is exact because every possible value of the parent must have a defined policy. This makes it suitable for variable elimination algorithms or sampling algorithms if the range of the parent is small and enough samples are generated.
A nearest neighbor decision policy.
A nearest neighbor decision policy. This policy computes an approximate decision from a sampling algorithm. The input to the class is an index (which holds (parent, decision) samples) a function that will combine a set of (decision, utility) samples into a single decision, and numNNSamples, the number of samples to use in a nearest neighbor algorithm. By default, this uses a VP-tree to store the samples.
Abstract class common to all multi-decision algorithms.
Abstract class common to all multi-decision algorithms. Multi-decision algorithms implement backward induction by 1) determining the order in which decisions can be computed 2) Implementing a single decision algorithm on each decision (in the proper order).
Note: Only OneTime algorithms are supported in multi-decision algorithms.
A multi-decision algorithm that uses Variable Elimination for each decision.
One-time Decision Metropolis-Hastings sampler.
A OneTime multi-decision algorithm that uses Importance sampling for each decision.
A OneTime multi-decision algorithm that uses Metropolis-Hastings sampling for each decision.
A OneTime multi-decision algorithm that uses Metropolis-Hastings sampling for each decision. A user must supple an instance of a ProposalMakerType, which indicates how to create a proposal scheme for each decision.
Trait for one time Decision Algorithms.
Decision VariableElimination algorithm that computes the expected utility of decision elements using the default elimination order.
Trait for Decision based Variable Elimination.
Trait for Decision based Variable Elimination. This implementation is hardcoded to use. Double utilities.
Trait that defines some common interface functions for decision algorithms. Every decision algorithm must define the function computeUtility().