In this modern age we are faced with the problem of a deluge of information. Information is accumulating at a faster rate than we can process it mainly due to technological advances that allow us to store and process more data than ever before. Thus the question of developing techniques to deal with this information explosion becomes highly significant. One such family of techniques are those which aim to exploit unlabeled data - precisely that data which is most readily available - and use it to improve performance in machine learning algorithms. This book presents several techniques which reflect the state of the art understanding of this emerging area of research.