Deep Learning & Data Efficiency
- Hao Dong, Imperial College London
- Time: 2019-04-04 14:00
- Host: Prof. Yizhou Wang
- Venue: Room 204, Courtyard No.5, Jingyuan
Deep learning is a data-driven technique, while in practice, data is difficult to be collected and labelled. Therefore, applying deep learning to practical problems required addressing the limited data problem. In this talk, I am going to talk about our two research directions for efficient use of data. The first one is to improve the deep learning performance while limited training data, we investigate three different types of data including signal, image and text. The second one is our recent work on generative adversarial networks (GAN), we learn to synthesise new images in unsupervised and semi-supervised ways, which successfully generate new data within limited training data.
Hao Dong is a final year PhD student at the Department of Computing of Imperial College London under the supervision of Prof. Yike Guo and Prof. Paul M. Matthews. He is interested in computer vision and deep learning and has publications on ICCV, TIFS, TMI, TNSRE, ACM MM and etc. He devotes to popularise Artifical Intelligence technology, he is the author of TensorLayer which won the Best Open Source Software Award at ACM Multimedia 2017.