Succinct Adaptive Manifold Transfer

Abstract

Capturing the relatedness of different domains is a key challenge in transferring knowledge across domains. In this paper, we propose an effective and efficient Gaussian process (GP) modelling framework, mTGP , that can explicitly model domain relatedness and adaptively control the space as well as the strength of knowledge transfer. mTGP takes both the discrepancy of input feature space and the discrepancy of predictive function into account in the transfer procedure. Specifically, mTGP adaptively selects a good latent manifold shared by different domains, and utilizes a parametric similarity coefficient to measure the predictive function covariance of different domains in this manifold. The latent shared manifold and the similarity coefficient are jointly learned in a coupled manner. By doing so, mTGP maximizes the strength of the shared knowledge transfer by choosing the transfer space with the best transfer capacity. More importantly, mTGP exploits a succinct and computationally efficient manifold learning approach so that it can be well trained with scarce target training data. Extensive experimental studies using 36 synthetic transfer tasks and 10 real-world transfer tasks show the effectiveness of mTGP on capturing the relatedness and the transfer adaptiveness. CCS CONCEPTS • Computing methodologies → Transfer learning.

Adaptive Transfer Kernel Learning for Transfer Gaussian Process Regression

“…It attracts increasing research attention in the last decade, and has achieved a great success in various real-world tasks [1], including sentiment analysis, digital classification, object recognition, etc. With research efforts largely confined to classification problems, the problem of transfer regression has been less studied despite its popularity in practical applications such as localization [2], time-series extrapolation [3], engineering design [4], motion detection [5], to name just a few.…”