内容来源：OpenScience

数值扰动对影像学影响的评估、建议、与解决方案

目前，可重复性已渐渐成为科学研究关注的重点之一。这一问题可能在不同领域、工具、数据集和计算基础设施中有不同的体现，但数值计算的不稳定性被认为是一个造成扰动的核心因素。在神经影像学中，操作系统，软件版本，处理流程都会对可重复性造成影响，而浮点运算的在其中的偏差还没有清晰的认识。Greg Kiar将会对数值扰动在影像学中的影响给出评估、建议、与解决方案。

报告信息

题目：Fuzzy environments for the perturbation evaluation and application of uncertainty quantification

报告语言：英文

分享嘉宾：Gregory Kiar, Research Scientist, Center for the Developing Brain, Child Mind Institute

Gregory Kiar, PhD, is a research scientist at the Child Mind Institute. Throughout his degrees in Biomedical Engineering, Greg has developed techniques to study biosignal data, a turn-key tool that allows researchers to generate maps of brain connectivity from diffusion MRI data, and techniques to assess and improve the stability of research in neuroscience.

Greg’s research bridges the fields of numerical analysis and brain imaging to evaluate and improve the trustworthiness of techniques used to study the brain. He has developed infrastructures that support the reliable evaluation of neuroimaging experiments, studied the role that unavoidable computing errors play in results, and developed techniques to consider such errors when building robust models of the brain. His work ultimately seeks to inform decision-making surrounding robust data collection, image processing and biomarker discovery. Greg has considerable experience in computational statistics, uncertainty quantification, pipeline development and machine learning.

In addition to his research aims, Greg has participated in the organization of over a dozen hackathon-style events focused on training and collaboration, and taught several full-length courses at the university level. Greg regularly takes on mentorship opportunities, which are often geared towards teaching academic writing, fundamental computational skills or research best practices.

摘要

With an increase in awareness regarding a troubling lack of reproducibility in analytical software tools, the degree of validity in scientific derivatives and their downstream results has become unclear. The nature of reproducibility issues may vary across domains, tools, data sets, and computational infrastructures, but numerical instabilities are thought to be a core contributor. In neuroimaging, unexpected deviations have been observed when varying operating systems, software implementations, or adding negligible quantities of noise. In the field of numerical analysis, these issues have recently been explored through Monte Carlo Arithmetic, a method involving the instrumentation of floating-point operations with probabilistic noise injections at a target precision. Exploring multiple simulations in this context allows the characterization of the result space for a given tool or operation. In this article, we compare various perturbation models to introduce instabilities within a typical neuroimaging pipeline, including (i) targeted noise, (ii) Monte Carlo Arithmetic, and (iii) operating system variation, to identify the significance and quality of their impact on the resulting derivatives. We demonstrate that even low-order models in neuroimaging such as the structural connectome estimation pipeline evaluated here are sensitive to numerical instabilities, suggesting that stability is a relevant axis upon which tools are compared, alongside more traditional criteria such as biological feasibility, computational efficiency, or, when possible, accuracy. Heterogeneity was observed across participants which clearly illustrates a strong interaction between the tool and data set being processed, requiring that the stability of a given tool be evaluated with respect to a given cohort. We identify use cases for each perturbation method tested, including quality assurance, pipeline error detection, and local sensitivity analysis, and make recommendations for the evaluation of stability in a practical and analytically focused setting. Identifying how these relationships and recommendations scale to higher order computational tools, distinct data sets, and their implication on biological feasibility remain exciting avenues for future work.

时间

• 北京时间[GMT+8] 7月8日(周五) 20:00~21:00
• 欧洲中部时间[CEST] 7月8日(周五) 14:00~15:00
• 美国东部时间[EST] 7月8日(周五) 08:00~09:00

Meeting ID: 9139 4010 836

报告流程：报告60分钟

主持人：徐婷  Child Mind Institute

其它：本次线上报告录屏会在结束后上传于B站:

OpenScience_CN  https://space.bilibili.com/252509184

---

组织团队(按名字首字母倒序排列)

OpenScience学术策划小组

张晗(博士), A*STAR, Singapore

张磊 (博士), University of Vienna, Austria

楊毓芳(博士), Freie Universität Berlin, Germany

杨金骉, MPI Psycholinguistics, the Netherlands

肖钦予, University of Vienna, Austria

王鑫迪(博士), 北京慧脑云

王庆(博士), Montreal Neurological Institute, Canada

金淑娴, Vrije Universiteit Amsterdam, the Netherlands

金海洋(博士), New York University Abu Dhabi, UAE

胡传鹏(博士), 南京师范大学

葛鉴桥 (博士), 北京大学

高梦宇(博士), 北京师范大学

耿海洋(博士), 香港大学

陈妍秀(博士), 中科院心理所

陈骥(博士), 浙江大学

neurochat团队

张文昊(UT Southwestern Medical Center, USA)

张洳源(上海交通大学)

张磊(University of Vienna, AUT)

应浩江(苏州大学)

徐婷(Child Mind Institute, USA)

王鑫迪(北京慧脑云)

滕相斌(MPI for Human Development, DEU)

鲁彬(中国科学院心理研究所)

孔祥祯(浙江大学)

胡传鹏(南京师范大学)

邸新(New Jersey Institute of Technology, USA)