arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2103.01636cs.AIcs.LGcs.MAcs.NE
收藏

用于可扩展和高效代理的稀疏训练理论

Sparse Training Theory for Scalable and Efficient Agents

University of Twente(特温特大学);Polytechnic Institute of Porto(波尔图理工学院);Eindhoven University of Technology(埃因霍温理工大学);Utrecht University(乌得勒支大学);University of Liège(列日大学)

Decebal Constantin Mocanu, Elena Mocanu, Tiago Pinto, Selima Curci, Phuong H. Nguyen, Madeleine Gibescu, Damien Ernst, Zita A. Vale

摘要:

人工智能的一个基本任务是学习。深度神经网络已被证明可以完美地应对所有学习范式,即监督学习、无监督学习和强化学习。然而,传统的深度学习方法利用云计算设施,并且不能很好地扩展到具有低计算资源的自主智能体。即使在云中,它们也会受到计算和内存的限制,并且它们不能用于为代理建模足够大的物理世界,这些代理假设具有数十亿个神经元的网络。这些问题在过去几年中通过新兴的稀疏训练主题得到了解决,稀疏训练从零开始训练稀疏网络。本文讨论了稀疏训练的最新技术水平、其挑战和局限性,同时介绍了几个新的理论研究方向,这些方向有可能缓解稀疏训练的局限性,推动深度学习的可扩展性远远超出其当前的界限。然而,理论的进步,在复杂的多代理设置的影响,从现实世界的角度进行了讨论,使用智能电网的案例研究。

英文摘要:

A fundamental task for artificial intelligence is learning. Deep Neural Networks have proven to cope perfectly with all learning paradigms, i.e. supervised, unsupervised, and reinforcement learning. Nevertheless, traditional deep learning approaches make use of cloud computing facilities and do not scale well to autonomous agents with low computational resources. Even in the cloud, they suffer from computational and memory limitations, and they cannot be used to model adequately large physical worlds for agents which assume networks with billions of neurons. These issues are addressed in the last few years by the emerging topic of sparse training, which trains sparse networks from scratch. This paper discusses sparse training state-of-the-art, its challenges and limitations while introducing a couple of new theoretical research directions which has the potential of alleviating sparse training limitations to push deep learning scalability well beyond its current boundaries. Nevertheless, the theoretical advancements impact in complex multi-agents settings is discussed from a real-world perspective, using the smart grid case study.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑