MASTER'S THESIS DEFENSE: James Duin
"Hierarchical Active Learning Application To Mitochondrial Disease Protein Dataset"
Committee Members: Dr. Stephen Scott (Advisor)
Dr. Juan Cui and Dr. Ashok Samal
Friday, April 14, 2017, 2:30 p.m.
256C AvH
Abstract:
This study investigates an application of active machine learning to a protein dataset developed to identify the source of mutations which give rise to mitochondrial disease. The dataset is labeled according to the protein’s location of origin in the cell; whether in the mitochondria or not, or a specific target location in the mitochondria’s outer or inner membrane, its matrix, or its ribosomes. This dataset forms a labeling hierarchy. A new machine learning approach is investigated to learn the high-level classifier, i.e., whether the protein is a mitochondrion, by separately learning finer-grained target compartment concepts and combining the results. This approach is termed active over-labeling. In experiments on the protein dataset it is shown that active over-labeling improves area under the precision-recall curve compared to standard passive or active learning. Because finer-grained labels are more costly to obtain, alternative strategies exploring using fixed proportions of a given budget to buy fine vs. coarse labels at various costs are compared and presented. Finally, we present a cost-sensitive active learner that uses a multi-armed bandit approach to dynamically choose the label granularity to purchase, and show that the bandit-based learner is robust to variations in both labeling cost and budget.
MASTER'S THESIS DEFENSE: Shi Cao
"Plant Image Processing: 3D Volume Reconstruction, Hyperspectral Information Mining and Visualization"
Committee Members: Dr. Hongfeng Yu (Advisor)
Dr. Sheng Wei and Dr. Lisong Xu
Tuesday, April 18 1:30 p.m.
347 Avery Hall
Abstract: Image processing techniques have been widely used in plant science for plant phenotyping studies. The fast algorithms are desired to process the massive image data. In this work, the traditional RGB color digital images from different taken angles are analyzed and we propose the algorithm to identify structures of plants by 3D volume reconstruction techniques. Another type of plant images which is known as hyperspectral images are studied. Analysis of hyperspectral images is of great importance in many scientific disciplines. Obtaining the spectral and spatial information simultaneously is a challenging task due to the high dimensionality of the hyperspectral images. We firstly develop a real-time interactive tool for exploring the hyperspectral images as a hyperspectral data cube. We present the strong correlation between information entropy and hyperspectral images with respect to the wavelength under which the hyperspectral images are taken. We design an information metric based transfer function allowing users to study the hyperspectral data cube by volume rendering techniques. In this manner, the transfer function dynamically changes with the regions of interest selected by users and both the spatial and spectral information are reserved. We show the usefulness of our approach in different scientific disciplines including plant science, physics and remote sensing. In addition, our transfer function also works for the traditional volumetric data and our method provides a new interactive tool for volume rendering.