Scholarlink

The training process of deep neural network (DNN) involves adjusting numerous parameters to minimize the prediction error. Several optimization algorithms, such as stochastic gradient descent (SGD), adaptive gradient algorithm (Adagrad), root mean squared propagation (RMSProp), adaptive moment estimation (Adam), and adaptive moment estimation with infinity norm (Adamax), among others, play a pivotal role in refining network weights and biases. Each algorithm exhibits unique characteristics, affecting convergence speed and final performance. The choice of optimizer often depends on the nature of the problem, dataset size, and network architecture. The present study assesses the performance of these optimizers based on convergence analysis.

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

Evaluating the Performance of Different Optimizers for Deep Learned Finite Elements

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ankit,
  • T. V. K. Subhash,
  • Dipjyoti Nath,
  • S. S. Gautam

摘要

The training process of deep neural network (DNN) involves adjusting numerous parameters to minimize the prediction error. Several optimization algorithms, such as stochastic gradient descent (SGD), adaptive gradient algorithm (Adagrad), root mean squared propagation (RMSProp), adaptive moment estimation (Adam), and adaptive moment estimation with infinity norm (Adamax), among others, play a pivotal role in refining network weights and biases. Each algorithm exhibits unique characteristics, affecting convergence speed and final performance. The choice of optimizer often depends on the nature of the problem, dataset size, and network architecture. The present study assesses the performance of these optimizers based on convergence analysis.