A Bayesian-learning technique for automatic pre-emptive loads through I/O devices via the mouse pointer

Abstract

In today's computing environment, it is well known that the computing bottleneck is rather at the I/O peripheral levels instead of at the level of CPU and memory. The access times to fetch data from an external device such as a CD-ROM, a network drive, or even the delay of dragging a mouse pointer to a desktop icon consumes seconds of time while CPU operations take nanoseconds. In this thesis, we show how our proposed Bayesian technique can anticipate certain memory intensive programs and how it can be used to preload its contents before the user selects the actual program. We evaluate the I/O peripheral of the mouse cursor and how to leverage historic mouse data to make these predictions. We show that using such Artificial Intelligence (AI) techniques results in a more productive computing environment relieving the user from waiting for a program to load.