TM 2010 - The 2nd International Workshop on Transfer Mining (TM'10)

Traditional data mining algorithms work well under a strict assumption: the training data and test data are drawn from the same distribution in the same feature space with the same set of class labels. In many real world problems, however, we usually do not have sufficient training data that satisfies this assumption. In order to apply traditional data mining algorithms, we may need to label a lot of training data, which is dreadfully expensive. It would be extremely useful if we can transfer the knowledge from other available data to our intended task while avoiding the effort of data labeling. In this workshop, we call for papers on the topic of transfer mining: transfer learning in data mining. There are several challenges to successfully transfer knowledge between different tasks. A first challenge is to judge the relatedness between tasks and avoid negative transfer. A second challenge is when given related tasks, decide what to transfer. Tasks may share some hyper-parameters, some features or some instances. It is nontrivial to decide what kind of knowledge should be transferred. Finally, how to transfer knowledge efficiently and effectively is another important issue. Transfer mining, which aims at transferring of knowledge between different domains and tasks in data mining, has emerged as one of the most active areas in data mining. There is a strong need to boost the research on knowledge transfer in the data mining community. Unlike in ICML/NIPS venues, the workshop will invite papers that address knowledge transfer from a data mining perspective. We welcome theoretical and applied disseminations that make efforts
to expose novel knowledge transfer methodology, frameworks and KDD processes for transfer mining.
to investigate effective (automated, human-machined-cooperated) principles and techniques for acquiring, representing, modeling and engaging transfer mining in real-world data mining.
to project trends and directions of transfer mining in both theories and applications.
With the success on the 1st International Workshop on Transfer Mining, the 2nd International Workshop on Transfer Mining will bring active researchers and industry practitioners together toward developing next-generation KDD theories. It will also further benefit the deployment of knowledge discovery in real world applications, and reduce the gap between data mining and machine learning, industry and practice.

Topics: (The topics of interest include but are not limited to the following)

Knowledge transfer on relational and heterogeneous data
Transfer mining for different types of data mining algorithms, including association rules, decision tree, KNN, K-means and so on.
Feature selection, extraction and construction in transfer mining.
Transferring among multiple related but different data sources.
Theory and algorithms to help avoid negative transfer.
Transfer mining on very large-scale data.
Transfer mining in high-impact applications, such as
Web applications, such as Web search, adverting, recommendation, social networking
Bioinformatics, medicine and biochemistry
Business, such as financial analysis and customer relationship management
NLP, such as sentiment analysis, parsing, machine translation, entity extraction
Wireless sensor networks and mobile computingg
Multimedia, such as image/video classification, multimedia tagging
Other novel applications