Numerical Optimization and Machine Learning

华为是电信基础设施和智能设备领域的全球领先者。该公司大量参与人工智能，并面临几个与机器学习技术优化相关的问题。该项目是华为加拿大公司与蒙雷阿尔理工学院四位教授的合作，他们在数值优化方面拥有多样化的专业知识。通过将这些专业知识应用到机器学习的工业应用中，其目的是促进数值优化方面的知识，开发和传播专门用于机器学习的优化工具，并培训意识到这些问题的高素质人员。在三年的时间里，该项目支持6名博士生、2名博士后研究员、14名本科生实习生和2名研究助理。监督遵循的是以研究人员为中心的等级结构，这种结构在过去被证明是成功的。算法及其相关理论是在GERAD研究中心开发的。它们是由行业合作伙伴推动的，但被设计为通用工具，可以使其他应用程序受益。代码和软件包由学生开发，并由两名专业研究助理进行验证。一旦足够成熟，他们将与华为的工程师合作在应用程序上进行测试。这项研究的预期结果是对现有优化方法的改进，更适合机器学习工业应用的新算法，以及更好的神经网络优化技术。新方法在科学出版物中进行了描述，软件包以其通用版本公开提供。该项目有助于蒙特雷亚尔在优化和机器学习两个领域的国际声誉，这两个领域对加拿大具有很高的战略意义。通过与一家在电信和人工智能领域处于领先地位的国际公司合作，研究人员获得了技术优势，使他们能够开发更先进的方法，并培训加拿大就业市场非常可取的人员。最后，该项目的所有方面都遵循公平、多样性和包容性的原则。

Huawei is a worldwide leader in telecommunication infrastructures and smart devices. The company is massively involved in AI and faces several issues related to the optimization in machine learning techniques. This project is a collaboration between Huawei Canada and four professors of Polytechnique Montréal with a diversified expertise in numerical optimization. By bringing this expertise into machine learning industrial applications, its aim is to advance knowledge on numerical optimization, to develop and disseminate optimization tools specialized for machine learning, and to train highly-qualified personnel aware of these issues. During three years, this project supports six PhD students, two postdoctoral fellows, fourteen undergraduate interns, and two research associates. Supervisions follow a hierarchical structure that revolves around the research associates and that has been proved successful in the past. Algorithms and their associated theory are developed at the GERAD research center. They are motivated by the industrial partner but are designed as generic tools that can benefit other applications. Codes and software packages are developed by students and validated by the two professional research associates. Once mature enough, they are tested on the applications in collaboration with Huawei's engineers. The anticipated outcomes of this research are the improvement of existing optimization methods, new algorithms that are better adapted to machine learning industrial applications, and better neural networks optimization techniques. The new methods are described in scientific publications and the software packages are made publicly available in their generic versions. This project contributes to the international reputation of Montréal in both fields of optimization and machine learning, which are highly strategic for Canada. By teaming up with an international company that is one of the leaders in telecommunications and in AI, researchers gain a technological edge that allows them to develop more advanced methods and train personnel that are highly desirable for the Canadian job market. Finally, all aspects of this project follow the principles of equity, diversity and inclusion.