Mathematics

Dr. Jeremy Eckhause
RAND Corporation

Many real world problems involve decisions that are made in sequence, where the consequence of one decision or policy affects subsequent decisions.  In this talk we describe a sequential decision model with the following structure: i) a set of decision epochs, ii) a set of system states, iii) a set of available actions with associated rewards or costs, and, possibly, iv) a set of state transition probabilities.  With the appropriate structure, such problems can be solved using dynamic programming, which is based on recursive computation.  Though solving models through dynamic programming may appear simple, it can be utilized across a wide range of applications and has generated significant mathematical theory.  This talk is intended as an introduction on the topic with some interactive exercises and illustrative examples