Redirigiendo al acceso original de articulo en 23 segundos...
Inicio  /  Algorithms  /  Vol: 15 Par: 4 (2022)  /  Artículo
ARTÍCULO
TITULO

Trinity: Neural Network Adaptive Distributed Parallel Training Method Based on Reinforcement Learning

Yan Zeng    
Jiyang Wu    
Jilin Zhang    
Yongjian Ren and Yunquan Zhang    

Resumen

Deep learning, with increasingly large datasets and complex neural networks, is widely used in computer vision and natural language processing. A resulting trend is to split and train large-scale neural network models across multiple devices in parallel, known as parallel model training. Existing parallel methods are mainly based on expert design, which is inefficient and requires specialized knowledge. Although automatically implemented parallel methods have been proposed to solve these problems, these methods only consider a single optimization aspect of run time. In this paper, we present Trinity, an adaptive distributed parallel training method based on reinforcement learning, to automate the search and tuning of parallel strategies. We build a multidimensional performance evaluation model and use proximal policy optimization to co-optimize multiple optimization aspects. Our experiment used the CIFAR10 and PTB datasets based on InceptionV3, NMT, NASNet and PNASNet models. Compared with Google?s Hierarchical method, Trinity achieves up to 5% reductions in runtime, communication, and memory overhead, and up to a 40% increase in parallel strategy search speeds.

 Artículos similares

       
 
Huihui Zhu, Hexiang Lin, Shaojun Wu, Wei Luo, Hui Zhang, Yuancheng Zhan, Xiaoting Wang, Aiqun Liu and Leong Chuan Kwek    
Integrated photonic chips leverage the recent developments in integrated circuit technology, along with the control and manipulation of light signals, to realize the integration of multiple optical components onto a single chip. By exploiting the power o... ver más
Revista: Information

 
Zeyu Xu, Wenbin Yu, Chengjun Zhang and Yadang Chen    
In the era of noisy intermediate-scale quantum (NISQ) computing, the synergistic collaboration between quantum and classical computing models has emerged as a promising solution for tackling complex computational challenges. Long short-term memory (LSTM)... ver más
Revista: Information

 
Ku Muhammad Naim Ku Khalif, Woo Chaw Seng, Alexander Gegov, Ahmad Syafadhli Abu Bakar and Nur Adibah Shahrul    
Convolutional Neural Networks (CNNs) have garnered significant utilisation within automated image classification systems. CNNs possess the ability to leverage the spatial and temporal correlations inherent in a dataset. This study delves into the use of ... ver más
Revista: Information

 
Zhenyu Feng, Qianqian You, Kun Chen, Houjin Song and Haoxuan Peng    
Evacuation simulation is an important method for studying and evaluating the safety of passenger evacuation, and the key lies in whether it can accurately predict personnel evacuation behavior in different environments. The existing models have good adap... ver más
Revista: Aerospace

 
Ping Huang and Yafeng Wu    
Airborne speech enhancement is always a major challenge for the security of airborne systems. Recently, multi-objective learning technology has become one of the mainstream methods of monaural speech enhancement. In this paper, we propose a novel multi-o... ver más
Revista: Aerospace