MSPA-MCS: Small-sample Network Inference in Computational Vision and Biology

PI: german, Donald J.机构：Johns Hopkins University标题：mmpa - mcs：计算视觉和生物学中的小样本网络推理摘要研究人员将开发理论模型和统计程序，旨在解决计算科学中的“小样本困境”。他们将专注于应用数学和计算机科学与分子生物学（生物信息学）和图像分析（计算机视觉）的界面上出现的问题。小样本困境（small-sample dilemma）是指原始数据的高维性（例如，传递基因和蛋白质状态或编码数字图像内容所需的大量变量）与可用于获取有用信息的相对较少的实际样本（重复、复制等）之间的差异。在生物信息学方面，研究人员将根据分子特征对细胞通路进行建模，并对包括疾病在内的表型进行分类。挑战在于从相对较少的训练示例中揭示复杂的分子相互作用网络。在计算机视觉方面，研究人员将研究携带视觉线索的图像特征之间的空间相互作用模式

Prop ID: CCF-0625687 PI: Geman, Donald J. Institution: Johns Hopkins University Title: MSPA-MCS: Small-sample Network Inference in Computational Vision and Biology AbstractThe investigators will develop theoretical models and statistical procedures which are designed to address the ``small-sample dilemma'' in the computational sciences. They will concentrate on problems arising at the interface of applied mathematics and computer science with both molecular biology (bioinformatics) and with image analysis (computer vision). The small-sample dilemma refers to the disparity between the high dimensionality of the raw data, for example the large number of variables necessary to convey the states of genes and proteins or to encode the content of digital images, and the relatively small number of actual samples (repetitions, replications, etc.) available for obtaining useful information. In bioinformatics, the investigators will work on modeling cellular pathways and classifying phenotypes, including disease, according to molecular signatures. The challenge is to uncover a complex network of molecular interactions from a relatively small number of training examples. In computer vision, the investigators will study patterns of spatial interactions among image features carrying visual cues