In its simplest form, the process of diagnosis is a decision-making process in which the diagnostician performs a sequence of tests culminating in a diagnostic decision. For example, a physician might perform a series of simple measurements (body tem- perature, weight, etc.) and laboratory measurements (white blood count, CT scan,...
This paper studies the problem of learning diagnostic policies from training examples. A diagnostic policy is a complete description of the decision-making actions of a diagnostician (i.e., tests followed by a diagnostic decision) for all possible combinations of test results. An optimal diagnostic policy is one that minimizes the expected...
A diagnostic policy species what test to perform next based on the results of previous tests and when to stop and make a diagnosis. Cost-sensitive diagnostic policies perform tradeoffs between (a) the costs of tests and (b) the costs of misdiagnoses. An optimal diagnostic policy minimizes the expected total cost....
This paper introduces the even-odd POMDP an approximation to POMDPs Partially Observable Markov Decision Problems in which the world is assumed to be fully observable every other time step. This approximation works well for problems with a delayed need to observe. The even-odd POMDP can be converted into an equivalent...
A common heuristic for solving Partially Observable Markov Decision Problems POMDPs is to first solve the underlying Markov Decision Process MDP and then construct a POMDP policy by performing a fixed depth lookahead search in the POMDP and evaluating the leaf nodes using the MDP value function. A problem with...
This paper addresses cost-sensitive classification in the setting where there are costs for measuring each attribute as well as costs for misclassification errors. We show how to formulate this as a Markov Decision Process in which the transition model is learned from the training data. Specifically we assume a set...
This paper introduces the even-odd POMDP, an approximation to POMDPs in which the world is assumed to be fully observable every other time step. The even-odd POMDP can be converted into an equivalent MDP, the
2MDP, whose value function, V*[subscript 2MDP], can be combined online with a 2-step lookahead search...
"The specifi c problem addressed in this proposal is the development of
good approximation algorithms for solving problems that have partial observability. The model we propose associates costs with obtaining information about the current state. We want to predict when and how much it is necessary to observe. We want...
The thesis focuses on model-based approximation methods for reinforcement
learning with large scale applications such as combinatorial optimization problems.
First, the thesis proposes two new model-based methods to stablize the
value–function approximation for reinforcement learning. The first one is the
BFBP algorithm, a batch-like reinforcement learning process which iterates between...