Xing, Eric ; Jebara, Tony
Transfer learning has received a lot of attention in the machine learning community over the last years, and several effective algorithms have been developed. However, relatively little is known about their theoretical properties, especially in the setting of lifelong learning, where the goal is to transfer information to tasks for which no data have been observed so far. In this work we study lifelong learning from a theoretical perspective. Our main result is a PAC-Bayesian generalization bound that offers a unified view on existing paradigms for transfer learning, such as the transfer of parameters or the transfer of low-dimensional representations. We also use the bound to derive two principled lifelong learning algorithms, and we show that these yield results comparable with existing methods.
991 - 999
ICML: International Conference on Machine Learning
2014-06-21 – 2014-06-26
Pentina A, Lampert C. A PAC-Bayesian bound for Lifelong Learning. In: Xing E, Jebara T, eds. Vol 32. Omnipress; 2014:991-999.
Pentina, A., & Lampert, C. (2014). A PAC-Bayesian bound for Lifelong Learning. In E. Xing & T. Jebara (Eds.) (Vol. 32, pp. 991–999). Presented at the ICML: International Conference on Machine Learning, Beijing, China: Omnipress.
Pentina, Anastasia, and Christoph Lampert. “A PAC-Bayesian Bound for Lifelong Learning.” edited by Eric Xing and Tony Jebara, 32:991–99. Omnipress, 2014.
A. Pentina and C. Lampert, “A PAC-Bayesian bound for Lifelong Learning,” presented at the ICML: International Conference on Machine Learning, Beijing, China, 2014, vol. 32, pp. 991–999.
Pentina A, Lampert C. 2014. A PAC-Bayesian bound for Lifelong Learning. ICML: International Conference on Machine Learning vol. 32. 991–999.
Pentina, Anastasia, and Christoph Lampert. A PAC-Bayesian Bound for Lifelong Learning. Edited by Eric Xing and Tony Jebara, vol. 32, Omnipress, 2014, pp. 991–99.
Link(s) to Main File(s)