arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2102.01732cs.LGcs.NE
收藏

大规模的真正稀疏神经网络

Truly Sparse Neural Networks at Scale

Eindhoven University of Technology(埃因霍温理工大学);University of Twente(特文特大学)

Selima Curci, Decebal Constantin Mocanu, Mykola Pechenizkiyi

摘要:

最近,稀疏训练方法已经开始被确立为人工神经网络中训练和推理效率的事实上的方法。然而,这种效率只是在理论上。在实践中,每个人都使用二进制掩码来模拟稀疏性,因为典型的深度学习软件和硬件都针对密集矩阵运算进行了优化。在本文中,我们采取正交的方法,我们表明,我们可以训练真正稀疏的神经网络收获他们的全部潜力。为了实现这一目标,我们介绍了三个新的贡献,专门为稀疏神经网络设计:(1)一个并行训练算法和相应的稀疏实现从头开始,(2)一个激活函数与非可训练的参数,有利于梯度流,和(3)一个隐藏的神经元的重要性度量,以消除冗余。总而言之,我们能够打破记录,并训练有史以来最大的神经网络的代表性力量-达到蝙蝠大脑的大小。结果表明,我们的方法具有最先进的性能,同时为环境友好的人工智能时代开辟了道路。

英文摘要:

Recently, sparse training methods have started to be established as a de facto approach for training and inference efficiency in artificial neural networks. Yet, this efficiency is just in theory. In practice, everyone uses a binary mask to simulate sparsity since the typical deep learning software and hardware are optimized for dense matrix operations. In this paper, we take an orthogonal approach, and we show that we can train truly sparse neural networks to harvest their full potential. To achieve this goal, we introduce three novel contributions, specially designed for sparse neural networks: (1) a parallel training algorithm and its corresponding sparse implementation from scratch, (2) an activation function with non-trainable parameters to favour the gradient flow, and (3) a hidden neurons importance metric to eliminate redundancies. All in one, we are able to break the record and to train the largest neural network ever trained in terms of representational power -- reaching the bat brain size. The results show that our approach has state-of-the-art performance while opening the path for an environmentally friendly artificial intelligence era.

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。

↑