Scholarlink

Since its birth in 1956, artificial intelligence (AI) has gone through multiple summers and winters. Driving by the significant advancements of neural network architectures, computing power and big data, the past decade has witnessed an unparalleled growth in machine learning (ML) and AI, leading to phenomenal breakthroughs in a wide spectrum of applications, e.g., speech recognition [191], image classification [141, 276], object detection [219, 323], etc. In fact, GPU throughput and memory have increased 10 \(\times \) in the last four years. By leveraging the parallelism of the GPU hardware and more training data, the transformer architecture can now train much more expressive models than ever, giving rise to a new era of foundation models. It is widely recognized that these intelligent applications will significantly enrich people’s lifestyle and improve human productivity.

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

Introduction to Continual and Reinforcement Learning for Edge AI

  • 章节

  • 发表日期：

    2025年

查看全文
  • Hang Wang,
  • Sen Lin,
  • Junshan Zhang

摘要

Since its birth in 1956, artificial intelligence (AI) has gone through multiple summers and winters. Driving by the significant advancements of neural network architectures, computing power and big data, the past decade has witnessed an unparalleled growth in machine learning (ML) and AI, leading to phenomenal breakthroughs in a wide spectrum of applications, e.g., speech recognition [191], image classification [141, 276], object detection [219, 323], etc. In fact, GPU throughput and memory have increased 10 \(\times \) in the last four years. By leveraging the parallelism of the GPU hardware and more training data, the transformer architecture can now train much more expressive models than ever, giving rise to a new era of foundation models. It is widely recognized that these intelligent applications will significantly enrich people’s lifestyle and improve human productivity.