Scholarlink

Software-related carbon dioxide emissions from the information and communications technology sector currently account for up to 3.9% of global emissions. With the increasing use of Machine Learning (ML) systems, this percentage of global emissions is estimated to grow. In this keynote, we embark on an interdisciplinary journey to explore the environmental sustainability of ML systems. Following a software engineering perspective, we see how to track and report green ML metrics in order to enable both their measurement and transparency. We then continue to optimize the carbon emissions and cost of ML systems during different stages of their lifecycle process by using green software tactics.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Environmental Sustainability of Machine Learning Systems: Reducing the Carbon Impact of Their Lifecycle Process

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Silverio Martínez-Fernández

摘要

Software-related carbon dioxide emissions from the information and communications technology sector currently account for up to 3.9% of global emissions. With the increasing use of Machine Learning (ML) systems, this percentage of global emissions is estimated to grow. In this keynote, we embark on an interdisciplinary journey to explore the environmental sustainability of ML systems. Following a software engineering perspective, we see how to track and report green ML metrics in order to enable both their measurement and transparency. We then continue to optimize the carbon emissions and cost of ML systems during different stages of their lifecycle process by using green software tactics.