University of Southern California
Ray R. Irani Hall
Molecular and Computational Biology
Computational Biology Colloquium
Chun-Nan Hsu
USC/Information Science Institute
"Applications of Machine Learning in Fluorescent Microscopic Cell Image Analysis”
Abstract:
Determining locations of protein expression is essential to understand protein function. Advances in green fluorescence protein (GFP) fusion proteins and automated fluorescence microscopy allow for rapid acquisition of large collections of protein localization images. Recognition of these cell images requires an automated image analysis system. The problem can be formulated as a multi-class classification problem in machine learning. In this talk, I will present a new learning algorithm, AdaBoost.ERC, coupled with weak and strong detectors, to improve the performance of automatic recognition of protein subcellular locations in cell images. AdaBoost.ERC outperforms other AdaBoost extensions. We demonstrate the benefit of weak detectors by showing significant performance improvements over classifiers using only strong detectors. We also empirically test our method's capability of generalizing to heterogeneous image collections. However, AdaBoost.ERC is a supervised learning method that requires sufficient training examples to achieve satisfying performance, but collecting training examples require skillful labeling by well-trained personnel which can be expensive. Semi-supervised learning has the potential of using a large set of unlabeled images for the recognition of subcellular organelle patterns, but the performance depends heavily on the selection of feature space to represent the image patterns. I will then present a feature space transformation method based on the spectral graph theory to improve semi-supervised learning. Experimental result shows that our feature space transformation method can improve the classification accuracy substantially. I will also briefly review some of my other works of applying machine learning in bioinformatics.
Thursday, March 4, 2010
2:00 pm
RRI 101
