Unsupervised and Transfer Learning: Challenges in Machine Learning, Volume 7

From the Foreword: This book is a result of an international challenge on Unsupervised and Transfer Learning (UTL) that culminated in a workshop of the same name at the ICML-2011 conference in Bellevue, Washington, on July 2, 2011; it captures the best of the challenge findings and the most recent research presented at the workshop. The book is targeted for machine learning researchers and data mining practitioners interested in "lifelong machine learning systems" that retain the knowledge from prior learning to create more accurate models for new learning problems. Such systems will be of fundamental importance to intelligent software agents and robotics in the 21st century. The articles include new theories and new theoretically grounded algorithms applied to practical problems. It addressed an audience of experienced researchers in the field as well as Masters and Doctoral students undertaking research in machine learning. The book is organized in three major sections that can be read independently of each other. The introductory chapter is a survey on the state of the art of the field of unsupervised and transfer learning providing an overview of the book articles. The first section includes papers related to theoretical advances in deep learning, model selection and clustering. The second section presents articles by the challenge winners. The final section consists of the best articles from the ICML-2011 workshop; covering various approaches to and applications of unsupervised and transfer learning.