modeling [6, 7] for 1-dimensional signals such as respiratory ones, or 3 to 4-dimensional

signals such as MR images [8].

 1.2 Of the Importance of Parameter Settings

 The complexity in using the gated experts approach stems from the number of

parameters that have to be set specifically for the particular data before the system can

work, and the potentially negative impact that one parameter change can have. Key

parameters can be divided in two groups: the global system parameters such as

performance measure and memory, which account for the accuracy in segmentation and

adaptation of the experts (most of which will be addressed in chapter 2), and the local

experts parameters, such as the kind of experts (linear vs. non-linear) or the type of basis

for non-linear experts (global or local), which determine the prediction by the experts

(they will be addressed in chapter 3).

 1.3 Outline

 This thesis is comprised of six chapters. Chapter 2 presents more in-depth

definitions concerning time series, stationarity, and the segmentation methods; chapter 3

focuses on describing data sets and problems of stationarity that entail generating a

synthetic test set, then presents the different models used for modeling the data. Chapter 4

covers the results of segmentation by the different methods on the test set. In chapter 5

real data is used, and the results are reviewed and compared. Finally, chapter 6

summarizes the results of the simulations analysis and conclusions, and then opens doors

for future work.