sequential_dp: Construct A Sequential Decision Problem
In bdpopt: Optimisation of Bayesian Decision Problems

Description Usage Arguments Details Value Author(s) See Also

Construct a sequential decision problem. The object constructed is just a list with components named as the arguments to the function.

1
2
3

sequential.dp(n.stages, post.sample, pred.sample, update.state,
term.decisions, term.obs.decisions, cont.decisions,
term.utility.fun, term.obs.utility.fun, cont.utility.fun)

`n.stages`	The number of stages in the sequential decision problem.
`post.sample`	A function taking a stage, a state and the simulation iteration count into a random sample from the posterior distribution of the model parameter given the value of the state when being at the given stage.
`pred.sample`	A function taking a stage, a list of parameter values and a decision into a list of random samples from the conditional distributions of the observable variable at the given stage. The i:th value in the list returned should be a sample from the predictive distribution conditioned on the i:th entry in the list of parameter values.
`update.state`	A function that takes a stage, a state, a decision `d` and a list of observations into a list of updated states. The updated values should be the ones obtained when combining the observations in order with the original state, given d.
`term.decisions`	A list the length of which must be equal to `n.stages`. The i:th element of the list specifies the terminal decisions available at stage i.
`term.obs.decisions`	A list the length of which must be equal to `n.stages`. The i:th element of the list specifies the terminal observation decisions available at stage i.
`cont.decisions`	A list the length of which must be equal to `n.stages`. The i:th element of the list specifies the continuation decisions available at stage i.
`term.utility.fun`	A list of terminal utility functions of length equal to `n.stages`. Each element of the list should be a function mapping a pair `(d, theta)` into to a numeric value, where `d` is a decision and `theta` is a parameter value.
`term.obs.utility.fun`	A list of terminal utility functions of length equal to `n.stages`. Each element of the list should be a function mapping a triple `(d, X, theta)` into to a numeric value, where `d` is a decision, `X` is an observation and `theta` is a parameter value.
`cont.utility.fun`	A list of terminal utility functions of length equal to `n.stages`. Each element of the list should be a function mapping a pair `(d, X)` into to a numeric value, where `d` is a decision and `X` is an observation.

For any stage i, at least one of the elements of the decision lists must be nonempty, i.e., the sum of length(term.decisions[[i]]), length(term.obs.decisions[[i]]) and length(cont.decisions[[i]] must be greater than or equal to 1.

For the last stage, all decisions must be terminal decisions, i.e., length(cont.decisions[[n.stages]]) must equal 0 and the sum of length(term.decisions[[n.stages]]) and
length(term.obs.decisions[[n.stages]]) must be greater than or equal to 1.

A list for which the components have the same names and are in the same order as the arguments to the function.

Sebastian Jobj<f6>rnsson jobjorns@chalmers.se

sequential.normal.dp

bdpopt documentation built on May 2, 2019, 9:18 a.m.