Time: 15:00 on 28/8/2015
Place: room 1002, building B1, HUST
Presenter: Dr. Nguyen Thi Thuy – Head, Dept. of Computer Science, Faculty of Information Technology, VNUA
Abstract:
Dictionary learning (DL) for sparse coding based classification has been widely researched in pattern recognition in recent years. Most of the DL approaches focused on the reconstruction performance and the discriminative capability of the learned dictionary. In talk, I will firstly present an overview of sparse coding problem and sparse coding based classification. I will then present a new method for learning discriminative dictionary for sparse representation, called Incoherent Fisher Discrimination Dictionary Learning (IFDDL), which combines the Fisher Discrimination Dictionary Learning (FDDL) and the Incoherent Dictionary Learning (IDL) method.
The sparse property has played an important role in the success of DL-based classification models. However, the sparsity constraint makes the learning problem expensive. Recently, there has been an emerged trend in relaxing the sparsity constraints (based on l0-norm and l1-norm) by using l2-norm. I will present some investigation into the relationship between the quality of the data and the dictionary learning issues that affect the performance of the classification system.
More about speaker: Dr. Nguyen Thi Thuy
Research interest:
– Computer vision: Object detection and recognition, semantic image segmentation, image classification.
– Machine learning: statistical learning models, Ensemble learning (Boosting, Random Forest); Sparse coding, Learning for classification.
– Applications