Computational Module

项目摘要 视觉研究已经加速发展，不仅需要了解视觉刺激是如何被接收的， 在大脑中处理，但对需要了解大脑如何解释，预测，和'理由' 关于它所看到的。机器学习/计算机视觉（ML/CV）方法现在是测试的基本工具 深度视觉/高阶视觉的理论，哈佛的研究人员率先开发了 这些算法使物体和人脸识别的基本发现。进一步发展这一 研究计划，视觉核心是推出一个计算模块，将支持开发 这一领域的新方法，并为寻求将ML/CV方法应用于其 research.该模块将配备一名ML/CV专家，他将生成新模型，创建管道， 训练模型，也为社区策划模型。视觉研究人员将特别受益于此 通过访问HMS高性能计算集群O2上的专用节点来获取资源， 由ML/CV专家管理。最后，计算模块将提供培训和教育 为愿景社区举办研讨会，以支持对这一极具影响力的项目感兴趣的下一代学员。 研究领域。该模块将由Gabriel Kreiman博士监督，他是使用和 视觉研究中ML/CV模型的开发，以及Vision Core的长期用户。通过生成 新的计算模块将为视觉社区提供一套可访问的专业知识和资源， 在视觉研究这一关键领域取得了进展。

Project Summary Vision research has accelerated toward a need to not only understand how visual stimuli are received and processed in the brain, but towards the need to understand how the brain interprets, predicts, and ‘reasons’ about what it is seeing. Machine learning/computer vision (ML/CV) approaches are now essential tools to test theories of deep vision/higher order vision, and investigators at Harvard have pioneered the development of these algorithms to make fundamental discoveries regarding object and face recognition. To further develop this research program, the Vision Core is launching a Computational Module that will support the development of new methods in this area, and to provide expertise to vision labs that seek to apply ML/CV methods to their research. The Module will be staffed by a ML/CV Specialist who will generate new models, create pipelines to train models, and also curate models for the community. Vision researchers will specifically benefit from this resource through access to a dedicated node on the HMS high-performance computing cluster, O2, which will be managed by the ML/CV specialist. Lastly, the Computational Module will provide training and educational workshops to the vision community to support the next generation of trainees interested in this highly impactful area of research. The Module will be supervised by Dr. Gabriel Kreiman, a leading expert in the use and development of ML/CV models in vision research, and a longstanding user of the Vision Core. By generating a set of accessible expertise and resources for the vision community, the new Computational Module will catalyze progress in this critical area of vision research.