Active learning based on a confidence measure to select new training data
Department of Electrical Engineering and Brain Science Research Center
Korea Advanced Institute of Science and Technology (KAIST)
DEI - Conference room
May 25th, 2012
Human utilizes active learning to develop their knowledge efficiently, and a new active learning model is presented and tested for automatic speech recognition systems. Human can self-evaluate their knowledge systems to identify the weak or uncertain topics, and seek for the answers by asking proper questions to experts (or teachers) or searching books (or webs). Then, the new knowledge is incorporated into the existing knowledge system.
Recently this active learning becomes very important to many practical applications such as speech recognition and text classification. On the internet abundant unlabelled data are available, but it is still difficult and time consuming to get accurate labels for the data. With the active learning paradigm, based on uncertainty analysis, a few data will be identified to be included in the training database and the corresponding labels will be asked to users. The answers will be incorporated into the current knowledgebase by an incremental learning algorithm. This process will be repeated to result in a high-accuracy classification system with minimum number of labelled training data.
The active learning algorithm had been applied to both a simple toy problem and a real-world speech recognition task. We introduced a uncertainty measure for each unlabelled data, which is calculated from the current classifier. The developed algorithm shows better recognition performance with less number of labelled data for the classifier training. In the future we will also incorporate a smooth transition on the selection strategy based on the exploitation-exploration trade-off. At the early stage of learning human utilizes exploitation while exploration is applied at the later stage.
Soo-Young Lee is Editor-in-Chief of the new INNS magazine, “Natural Intelligence”, and has served as Publication Committee Chair. He is a Past-President of Asia-Pacific Neural Network Assembly (APNNA). He received Leadership Award (1994) and Presidential Award (2001) from INNS, Excellent Service Award (2004) and Outstanding Achievement Award (2009) from APNNA, and Biomedical Wellness Award (2008) and ICA Unsupervised Learning Pioneer Award (2010), respectively. He also has served as General Chair and Program Chair for many international conferences.
He had emphasized interdisciplinary research and education. He has served as Director of Brain Neuroinformatucs Research Program (1998-2008) to lead a multidisciplinary research team (about 30 professors from neuroscience to engineering) for understanding neural information processing mechanism and developing brain-inspired intelligent machine, i.e., Artificial Brain. In 2002 he had also served as first Chair of BioSystems Department at KAIST to educate young students with both biological science and information technology background.
After graduating with MS in Korea, he had graduated from Polytechnic Institute of New York with PhD in 1984. After a short stay for an industrial company in US, he started academic career at KAIST, Korea. He understands international multi-culture and had sabbatical leaves in US, Germany, and Japan.
Artificial intelligence, robotics and computer vision