Scholarlink

Abstract <p>Non-orthogonal multiple access is a strategy for improving resource sharing among users in 5G and beyond wireless communication networks. However, establishing equitable power allocation in NOMA systems is a substantial difficulty that directly impacts system performance. In this context, BiLSTM-based deep learning algorithms provide an efficient approach for optimizing power allocation. This research employs bidirectional long short-term memory networks to improve power allocation in NOMA systems. The research improved the BiLSTM model’s hyperparameter (number of epochs, batch size, and optimizers) to increase system performance. It has been tested at various signal-to-noise ratio levels, with performance metrics including bit error rate, symbol error rate, throughput, Jain index and proportional fairness index. Research results indicate that the BiLSTM model produced a BER value of <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(8.93e^{-9}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>8.93</mn> <msup> <mi>e</mi> <mrow> <mo>-</mo> <mn>9</mn> </mrow> </msup> </mrow> </math></EquationSource> </InlineEquation> with 50 epochs and a batch size of 64, resulting in an accuracy of 99.9999991%, by optimizing power allocation with the appropriate hyperparameter tuning using grid search and Bayesian-based algorithm and utilizing the Adam optimizer. Furthermore, the model has been shown to be substantially more efficient than traditional techniques, with a training time of 38.51&#xa0;s and memory consumption of 962.63 MB. These findings show that the suggested technique not only achieves excellent accuracy but also has a considerable benefit in real-world applications due to shorter processing times and lower memory usage. It also illustrates that this model is an effective tool for improving performance in NOMA systems. The paper attempts to solve gaps in the current literature by using BiLSTM-based power allocation optimization in NOMA systems. In the future, it is proposed that the model be evaluated with bigger datasets and used in a variety of communication contexts.</p>

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

Fair Power Allocation in NOMA Systems: BiLSTM-Based Hyperparameter Optimization

  • 期刊论文

  • 发表日期：

    2025年10月15日

查看全文
  • Seda Kirtay,
  • Veysel Gokhan Bocekci,
  • Kazim Yildiz

摘要

Abstract

Non-orthogonal multiple access is a strategy for improving resource sharing among users in 5G and beyond wireless communication networks. However, establishing equitable power allocation in NOMA systems is a substantial difficulty that directly impacts system performance. In this context, BiLSTM-based deep learning algorithms provide an efficient approach for optimizing power allocation. This research employs bidirectional long short-term memory networks to improve power allocation in NOMA systems. The research improved the BiLSTM model’s hyperparameter (number of epochs, batch size, and optimizers) to increase system performance. It has been tested at various signal-to-noise ratio levels, with performance metrics including bit error rate, symbol error rate, throughput, Jain index and proportional fairness index. Research results indicate that the BiLSTM model produced a BER value of \(8.93e^{-9}\) 8.93 e - 9 with 50 epochs and a batch size of 64, resulting in an accuracy of 99.9999991%, by optimizing power allocation with the appropriate hyperparameter tuning using grid search and Bayesian-based algorithm and utilizing the Adam optimizer. Furthermore, the model has been shown to be substantially more efficient than traditional techniques, with a training time of 38.51 s and memory consumption of 962.63 MB. These findings show that the suggested technique not only achieves excellent accuracy but also has a considerable benefit in real-world applications due to shorter processing times and lower memory usage. It also illustrates that this model is an effective tool for improving performance in NOMA systems. The paper attempts to solve gaps in the current literature by using BiLSTM-based power allocation optimization in NOMA systems. In the future, it is proposed that the model be evaluated with bigger datasets and used in a variety of communication contexts.