首页> 外文会议>IEEE MTT-S International Microwave Symposium >Parallel matrix neural network training on cluster systems for dynamic FET modeling from large datasets
【24h】

Parallel matrix neural network training on cluster systems for dynamic FET modeling from large datasets

机译:在大型数据集上进行动态FET建模的集群系统上的并行矩阵神经网络训练

获取原文

摘要

This paper presents a powerful and general parallel artificial neural network training technique with parallel computing on cluster systems. Large numbers of training samples are distributed to multiple computers on a cluster system to achieve a high speed-up for training. The method is evaluated with respect to two examples of an advanced dynamic nonlinear simulation model for GaN transistors where the training set is measured large-signal waveform data from an NVNA. For advanced models in a high dimensional space with large training sample size, the proposed approach is demonstrated to reduce the total training time by a factor of 35.
机译:本文提出了一种功能强大且通用的,在集群系统上具有并行计算能力的并行人工神经网络训练技术。大量的训练样本被分发到群集系统上的多台计算机上,以加快训练速度。该方法是针对GaN晶体管的高级动态非线性仿真模型的两个示例进行评估的,其中,训练集是从NVNA测得的大信号波形数据。对于高维度空间中具有大量训练样本量的高级模型,所提出的方法被证明可以将总训练时间减少35倍。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号