Poor coding practices in ML increase carbon emissions, study finds

A new research paper from arXiv investigates the environmental impact of inefficient coding practices in machine learning applications, specifically focusing on TensorFlow and Keras. The study quantifies how resource leaks, such as improper model reuse and unreleased tensor references, lead to increased energy consumption and carbon emissions. Preliminary results indicate that these coding smells can elevate electricity usage by approximately 32% to 46%, highlighting the need for better resource lifecycle management in ML development. AI

IMPACT Highlights the significant environmental cost of inefficient ML code, urging developers to prioritize sustainability.

RANK_REASON Academic paper detailing empirical investigation of resource leaks in ML code and their environmental impact. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Improper Model Reuse
• Keras
• TensorFlow
• Unreleased Tensor References

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →