[Summary] A Survey on Transfer Learning

Topic: A Survey on Transfer Learning

Author: S. J. Pan and Q. Yang

Summary:

Many machine learning methods work well only under the assumption of "the training and test data are drawn from the same feature space and the same distribution". When the distribution changes, most models need to be rebuilt, and it is expensive or impossible to recollect the needed training data and rebuild the models. It would be nice to reduce the need and effort to recollect the training data. In such cases, knowledge transfer or transfer learning between task domains would be desirable.

In this survey, a domain D consists of two components: a feature space X and a marginal probability distribution P(X). In general, if two domains are different, then they may have different feature spaces or different marginal probability distributions.
Given a specific domain, D={X, P(X)}, a task consists of two components: a label space Y and an objective predictive function f(.), which is not observed but can be learned from the training data. From a probabilistic viewpoint, f(x) can be written as P(y|x).
For simplicity, in this survey, we only consider the case where there is one source domain DS, and one target domain, DT.



In transfer learning, we have the following three main research issues:

1. What to transfer: which part of knowledge can be transferred across domains or tasks.
2. How to transfer: learning algorithms need to be developed to transfer the knowledge
3. When to transfer: in which situations, transferring skills should be done.

The transfer learning can be categorized into three sets:

• Inductive transfer learning
• Transductive transfer learning
• Unsupervised transfer learning

Inductive transfer learning

Given a source domain DS and a learning task TS, a target domain DT and a learning task TT , inductive transfer learning aims to help improve the learning of the target predictive function fT(.) in DT using the knowledge in DS and TS, where TS/=TT .
In this case, some labeled data in the target domain are required to induce an objective predictive model fT(.) for use in the target domain. It can be categorized into two cases:

• A lot of labeled data in the source domain are available. In this case, the inductive transfer
  learning setting is similar to the multitask learning setting.
• No labeled data in the source domain are available. In this case, the inductive transfer
  learning setting is similar to the self-taught learning setting.

Transductive transfer learning

Given a source domain DS and a corresponding learning task TS, a target domain DT and a corresponding learning task TT , transductive transfer learning aims to improve the learning of the target predictive function fT(.) in DT using the knowledge in DS and TS, where DS/=DT and TS/=TT. In addition, some unlabeled target-domain data must be available at training time. It can be categorized into two cases:

• The feature spaces between the source and target domains are different, XS/=XT.
• The feature spaces between domains are the same, XS/=XT , but the marginal probability distributions of the input data are different, P(XS)/=P(XT).

 

Unsupervised transfer learning

Given a source domain DS with a learning task TS, a target domain DT and a corresponding learning task TT, unsupervised transfer learning aims to help improve the learning of the target predictive function fT(.) in DT using the knowledge in DS and TS, where TS/=TT and YS and YT are not observable.
In this case, there are no labeled data available in both source and target domains in training.

So far, transfer learning techniques have been mainly applied to small scale applications with a limited variety, such as sensor-network-based localization, text classification, and image classification problems. In the future, transfer learning techniques will be widely used to solve other challenging applications, such as video classification, social network analysis, and logical inference.