Title: | A Low Rank Multivariate General Linear Model of Multi-Subject fMRI Data and a Non-convex Optimization Algorithm for Brain Response Comparison |
Speaker: | Dr. Tingting Zhang, Department of Statistics, University of Virginia, U.S.A. |
Time/Place: | 16:00 - 17:00 FSC1217, Fong Shu Chuen Library, HSH Campus, Hong Kong Baptist University |
Abstract: | This talk focuses on comparing brain responses to different stimuli and identifying brain regions with different responses using multi-subject, stimulus-evoked functional magnetic resonance imaging (fMRI) data. To jointly model spatially distributed voxels’ brain responses to designed stimuli, we present a new low-rank multivariate general linear model (LRMGLM) for stimulus-evoked fMRI data. The new model not only accommodates the variation of the brain activity across different regions and stimulus types, but also enables information “borrowing” across voxels and uses much fewer parameters than typical nonparametric models for fMRI data. To estimate the proposed LRMGLM, we introduce a new penalized optimization function, which incorporates the spatial information of the voxels, the temporal smoothness of brain activity, and imposes a sparsity constraint on the identified brain regions. We develop an efficient optimization algorithm to minimize the optimization function and identify the regions with different responses to stimuli. We show that the proposed method can outperform several existing voxel-wise methods by achieving both high sensitivity and specificity. We apply the proposed method to the fMRI data collected in an emotion study, comparing brain responses to different emotional stimuli. |
Title: | Inverse Problems in Adaptive Optics |
Speaker: | Prof. Ronny Ramlau, Industrial Mathematics Institute Johannes Kepler University Linz / Johann Radon Institute for Computational and Applied Mathematics (RICAM) Austrian Academy of Sciences (AAS), Austria |
Time/Place: | 16:00 - 17:00 FSC1217, Fong Shu Chuen Library, HSH Campus, Hong Kong Baptist University |
Abstract: | "Currently there is a new generation of large astronomical telescopes under construction, e.g. the European Extremely Large Telescope (E-ELT) of the European Southern Observatory (ESO) with a mirror diameter of 39 meters or the Thirty Meter Telescope (TMT), build by a consortium headed by Caltech. The operation of those huge telescopes require new mathematical methods in particular for the Adaptive Optics systems of the telescopes. The image quality of ground based astronomical telescopes suffers from turbulences in the atmosphere. Adaptive Optics (AO) systems use wavefront sensor measurements of incoming light from guide stars to determine an optimal shape of deformable mirrors (DM) such that the image of the scientific object is corrected after reflection on the DM(s). The solution of this task involves several inverse problems: First, the incoming wavefronts have to be reconstructed from wavefront sensor measurements. The next step involves the solution of the Atmospheric Tomography problem, i.e., the reconstruction of the turbulence profile in the atmosphere. Finally, the optimal shape of the mirrors has to be determined. As the atmosphere changes frequently, these computations have to be done in real time. In the talk we introduce mathematical models for the elements of different Adaptive Optics system such as Single Conjugate Adaptive Optics (SCAO) or Multi Conjugate Adaptive Optics (MCAO). In particular, we consider the reconstruction of the incoming wavefront from Shack-Hartmann and Pyramid sensor data and present different reconstruction approaches for the atmospheric tomography problem. The numerical results for each of the sub-tasks confirm that the methods achieve the accuracy and speed required for the operation on ELTs. Finally we will present first results for the reconstruction of the point spread function of the telescope based on Adaptive Optics data and the related deblurring of astronomical images. " |
Title: | Automated Model Building and Deep Learning |
Speaker: | Prof. Wang Xiao, Department of Statistics, Purdue University, U.S.A. |
Time/Place: | 15:30 - 16:30 FSC1217, Fong Shu Chuen Library, HSH Campus, Hong Kong Baptist University |
Abstract: | Analysis of big data demands computer aided or even automated model building. It becomes extremely difficult to analyze such data with traditional statistical models and model building methods. Deep learning has proved to be successful for a variety of challenging problems such as AlphaGo, driverless cars, and image classification. Understanding deep learning has apparently limited, which makes it difficult to be fully developed. In this talk, we focus on neural network models with one hidden layers. We provide an understanding of deep learning from an automated modeling perspective. This understanding leads to a sequential method of constructing deep learning models. This method is also adaptive to unknown underlying model structure. This is a joint work with Chuanghai Liu. |
We organize conferences and workshops every year. Hope we can see you in future.
Learn MoreProf. M. Cheng, Dr. Y. S. Hon, Dr. K. F. Lam, Prof. L. Ling, Dr. T. Tong and Prof. L. Zhu have been awarded research grants by Hong Kong Research Grant Council (RGC) — congratulations!
Learn MoreFollow HKBU Math