2000 character limit reached
Towards the full information chain theory: expected loss and information relevance (1301.2020v1)
Published 10 Jan 2013 in physics.data-an, cs.IT, and math.IT
Abstract: When additional information sources are available, an important question for an agent solving a certain problem is how to optimally use the information the sources are capable of providing. A framework that relates information accuracy on the source side to information relevance on the problem side is proposed. An optimal information acquisition problem is formulated as that of question selection to maximize the loss reduction for the problem solved by the agent. A duality relationship between pseudoenergy (accuracy related) quantities on the source side and loss (relevance related) quantities on the problem side is observed.