Research Activities – Workshop ISE

12. Workshop ISE – Fri 19/11 11B24 2:30-3:30

Topic: Relation Learning – A New Approach to Face Recognition
Abstract: Most of current machine learning methods used in pattern recognition systems require sufficient data to build pattern models or data descriptions. However data insufficiency is currently an issue. This talk presents a new learning method to tackle this issue. The proposed learning method employs relations between training samples to build relational pattern models. Preliminary experiments performed on the AT&T and FERET face corpora show a good improvement for face recognition when small facial data sets are available for training.
Presenter: Len Bui, PhD student
Date: Friday 19 November 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24

11. Workshop ISE – Fri 12/11 11B24 2:30-3:30

Topic: CLONALG for classification
Abstract: Currently, the majority of artificial immune systems (AIS) encompass two different types of immune inspired algorithms, namely: negative selection and clonal selection. Clonal selection principal is a form of evolutionary process where only those immune cells which specifically recognize the antigens are proliferated (cloned) and then mutated. Some immune cells that can easily bind antigens are kept as memory cells to avoid relearning if infected by antigens of similar pattern in future. This also results in faster immune response by already having high affinity antibodies present in the immune system. Inspired from clonal selection paradigm, CLONALG algorithm has been developed for pattern recognition, however, it has not been applied to this task. I will talk about some additional features that can be incorporated into CLONALG algorithm so that it can be applied for pattern recognition, namely classification problems.
Presenter: Anurag Sharma, PhD student
Date: Friday 12 November 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24

10. Workshop ISE – Fri 5/11 11B24 2:30-3:30

Topic: Multi-Sphere Support Vector Clustering
Abstract: Current support vector clustering determines the smallest sphere that encloses the image of a dataset in feature space. This sphere when mapped back to data space will form a set of contours that can be interpreted as cluster boundaries for the dataset. However this method does not guarantee that the single sphere and the resulting cluster boundaries can best describe the dataset if there are some distinctive data distributions in this dataset. We propose multi-sphere support vector clustering to address this issue. Data points in data space are mapped to a high dimensional feature space and a set of smallest spheres that encloses the image of the data is determined. This set of spheres when mapped back to data space will form a set of contours that can be interpreted as cluster boundaries. Experiments on several datasets are performed to demonstrate that the proposed approach provides a better cluster analysis than the current support vector clustering and other popular clustering methods.
Presenter: Trung Le, PhD student
Date: Friday 5 November 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24

9. Workshop ISE – Fri 29/10 11B24 2:30-3:30

Topic: Negative Selection, Its Characteristics, Problems, and Potential New Application Areas
Abstract: Negative selection is a branch of Artificial Immune Systems (AIS), which are inspired by the observation of the behaviors and the interaction of antibodies and antigens in a biological system. It mimics the way a human body detects and destroys harmful antigens. It is neat yet powerful and has found a wide range of applications, especially in abnormal detection.
I will talk about negative selection, its unique characteristics, its problems, and potential new application areas.
Presenter: Dr Wanli Ma
Date: Friday 29 October 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24

8. Workshop ISE – Fri 22/10 11B24 2:30-3:30

Topic: Indications of chaotic properties in time series of brain electrical activity
Abstract: Continuing from my last workshop, I will present some preliminary issues from applying chaos theory to BCI research. Chaotic properties of brain electrical activity from different recording regions and from different brain states are compared and the results show that there are strong indications of chaotic properties for seizure activity, but very weak or even no indication for other activities. The nonlinear prediction error and an estimate of an effective correlation dimension are used in combination with the method of iterative amplitude adjusted surrogate data to analyse sets of EEG time series.
Presenter: Tuan Hoang, PhD student
Date: Friday 22 October 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24

7. Workshop ISE – Fri 15/10 11B24 2:30-3:30

Topic: Age, Gender, and Affect Classification Methods in the INTERSPEECH 2010 Paralinguistic Challenge
Abstract: The INTERSPEECH 2010 Paralinguistic Challenge is a Special Session of the INTERSPEECH Conference held in Makuhari, Tokyo, Japan 26-30 Sep 2010. The INTERSPEECH 2010 Paralinguistic Challenge addresses three selected sub-challenges: Age Sub-Challenge (to determine the age of speakers); Gender Sub-Challenge (to solve a two-class classification task) and Affect Sub-Challenge (to ask for determination of speakers' interest in ordinal representation). There are 11 papers selected for this session.
Feature extraction and classification methods as well as experimental results in the 11 selected papers will be presented.
The paper list can be found at http://www.interspeech2010.org/program/session.php?id=5510  
Presenter: Phuoc Nguyen, Master by Research student
Date: Friday 15 October 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24

6. Workshop ISE – Fri 8/10 11B24 2:30-3:30

Topic: Anomaly Detection and One Class Support Vector Machine
Abstract: Anomaly detection refers to detecting patterns in a given data set that do not conform to an established normal behavior. It is widely applied in a variety of domains such as computer security, network intrusion detection, machine fault detection and banking service. In this presentation, I will review popular anomaly detection techniques then focus on One Class Support Vector Machine and propose some possible extensions to this technique. Research questions discussed during the presentation include: 1) What are anomaly detection and novelty detection? 2) What is the difference between anomaly detection and outlier detection? 3) What is taxonomy of anomaly detection? 4) How is One Class Support Vector Machine developed from Support Vector Machine? 5) How can One Class Support Vector Machine be used to deal with various data distributions?
Presenter: Trung Le, PhD student
Date: Friday 8 October 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24

5. Workshop ISE – Fri 1/10 11B24 2:30-3:30

Topic: Real-time face detection
Abstract: Face detection is a two-class classification problem and has broad practical applications including user interfaces, image databases, and teleconferencing. Some well-known methods such as Principal Component Analysis (PCA), Multi-Layer Perception (MLP) and Support Vector Machines (SVM) can provide low detection error rates in face detection; however they cannot be used in real-time face detection systems. In this seminar, I will present a new approach proposed by Viola and Jones. The proposed face detection system is most clearly distinguished from previous approaches in its ability to detect faces extremely rapidly.
Presenter: Len Bui, PhD student
Date: Friday 1 October 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24

4. Workshop ISE – Fri 24/9 11B24 2:30-3:30

Topic: Chaos Theory and Its Application to EEG-Based Brain Computer Interfaces
Abstract: Brain-Computer Interface (BCI) is an emerging research field attracting a lot of research effort from researchers around the world. Its aim is to build a new communication channel that allows a person to send commands to an electronic device using his/her brain activities. The majority of promising non-invasive BCI systems to date exploit electroencephalography (EEG) signal, mainly due to its fine temporal resolution, ease of use, portability and low set-up cost. From various experiments, it has been well established that brain activities and EEG recordings show many characteristics of chaotic behaviour and could be explained on the basis of mathematical chaos theory. In this seminar, I will present some core concepts of chaos theory and its application to EEG-based BCI systems
Presenter: Tuan Hoang, PhD student
Date: Friday 24/9 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24

3. Workshop ISE – Fri 22/10 11B24 2:30-3:30

Topic:
Abstract:
Presenter: Tuan Hoang, PhD student
Date: Friday 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24

2. Workshop ISE – Fri 22/10 11B24 2:30-3:30

Topic:
Abstract:
Presenter: , PhD student
Date: Friday 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24

1. Workshop ISE – Fri 22/10 11B24 2:30-3:30

Topic:
Abstract:
Presenter: , PhD student
Date: Friday 2010
Time: 2.30 PM – 3:30 PM
Location: 11B24