34、Building Sustainable Enterprise - Grade AI Platforms: A Comprehensive Guide

异步汪仔

于 2025-09-03 13:11:35 发布

阅读量14

点赞数

CC 4.0 BY-SA版权

分类专栏：构建负责任的人工智能文章标签： Federated Learning Energy Efficiency CO₂e Emissions

本文链接：https://blog.youkuaiyun.com/d3e4f/article/details/151338573

构建负责任的人工智能专栏收录该内容

40 篇文章 ¥499.90

订阅专栏¥69.90

会员秒杀 ¥9.9 重磅福利

超级会员免费看

Building Sustainable Enterprise - Grade AI Platforms: A Comprehensive Guide

1. Understanding the Basics of Federated Learning and Energy Efficiency

Federated learning (FL) has emerged as a promising approach in AI, but its energy efficiency is a crucial concern. Non - IID datasets in FL require more training rounds compared to IID datasets, which converge faster.

The FedADAM optimizer has shown effectiveness in certain studies. For CIFAR - 10 and SpeechCmd datasets, it outperformed FedAvg. When using five local epochs (LEs), FedAdam had lower CO₂e emissions than centralized learning and FL with FedAvg. However, in ImageNet experiments with non - IID data, FedAdam had higher emissions due to the unbalanced dataset, which led to longer training times. </