Part II - Self-Organization applied to Recommender Systems

Progress isn't made by early risers.
It's made by lazy men trying to find
easier ways to do something.

Robert Heinlein

This part consists of three chapters. The first two are theoretical, giving the necessary background for chapter three, in which we use the knowledge from the previous chapters in our attempt to implement and evaluate recommender systems based on the SOM paradigm.

The purpose of the first chapter is to give the reader a basic understanding in what artificial neural networks are. Motivations behind the development of artifical neural networks are discussed, the basic biological neuron and the artificial neuron are described, network architectures and learning processes are outlined, and a brief historical summary of some of the main events in the field of artifical neural networks is given.

The second chapter deals with self-organizing systems in which Kohonen's SOM is the main topic. Self-organization as a biological phenomena is discussed and the requirements for self-organization are described. Then, after a short introduction to the SOM paradigm, the incremental SOM algorithm is outlined in detail together with a presentation of how parameters for the algorithm should be selected, how to measure the quality of the SOM visualization and clustering issues in the SOM context. Some important properties regarding the final SOM is also presented and a discussion of the theoretical aspects of the SOM is given. Presentations of related work in the area of using the SOM paradigm in recommender systems are given at the end of the chapter.

Finally, in chapter three, we describes why we have chosen to use the SOM paradigm as a part of a recommender system by referring back to Part I and the issues regarding recommender systems that was presented there. We proceed by showing how we have implemented different recommender systems based on the SOM paradigm. This is done with pseudo code and descriptive commentaries of the code. The systems are then evaluated in the same manner as in Part I.