Title: Occam' s Razor : Understanding the Cost of Complexity Abstract: In this talk I will provide a brief overview of various ways that the dictum "all else being equal, simpler models are better" can be understood in the context of machine learning and the theory of prediction. I will begin by demonstrating that this idea is not necessarily true if we do not choose an appropriate definition of simplicity. I will then discuss some of the ways in which simplicity can be quantitatively measured, and show how these measurements are involved when analyzing the performance of learning algorithms. The talk will be focussed on providing insight, rather than on understanding the results in rigorous detail.