Variational Inference-based Deep Kernel Learning for Few-shot Incremental Learning.

July 28th, 2022

Thesis Type Bachelor
Status Open
Supervisor(s) Stefan Decker
Advisor(s) Yongli Mou

In few-shot incremental learning, the challenges are two-fold, generalizing from a small number of training data for the novel classes and avoiding catastrophic forgetting of the previously encountered classes. Combining Gaussian process (GPs) with deep learning methods via deep kernel learning (DKL), GP-tree (proposed by Achituve et al.) is especially compelling due to the strong representational power induced by the network and a natural fit for this problem. The kernel is a vital component here for the further classification task, therefore, training a good deep kernel is essential.

The objective of this thesis is to adapt variational methods to improve the performance of kernel learning in few-shot incremental learning.

If you are interested in this thesis, do not hesitate to contact us via

Please find the seed literature as followings:

  1. Achituve I, Navon A, Yemini Y, Chechik G, Fetaya E. Gp-tree: A gaussian process classifier for few-shot incremental learning. InInternational Conference on Machine Learning 2021 Jul 1 (pp. 54-65). PMLR.
  2. Wilson AG, Hu Z, Salakhutdinov R, Xing EP. Deep kernel learning. InArtificial intelligence and statistics 2016 May 2 (pp. 370-378). PMLR.


Knowledge about Machine Learning
Programming language – Python
Deep Learning Framework – PyTorch