研究目的
To solve the issues of conventional network planning approach by applying machine learning for data-driven model development and transmission quality prediction, focusing on neural network based modulation format classification.
研究成果
Machine learning, particularly MLP classifiers, can effectively predict physical layer modulation formats with high accuracy, offering a dynamic and scalable solution for next-generation network resource optimization. The strongest classifier achieved ~98.4% accuracy with a training time of ~20s.
研究不足
The study focuses on modulation format classification using synthetic data generated by GN simulations, which may not fully capture all real-world transmission scenarios. Future work includes dataset enrichment through in-field data.
1：Experimental Design and Method Selection:
The study employs multi-layer perceptron (MLP) architectures for modulation format classification based on features like symbol rate, channel load, number of spans, etc.
2：Sample Selection and Data Sources:
The dataset was generated using Gaussian noise (GN) simulations, consisting of training, validation, and test datasets.
3：List of Experimental Equipment and Materials:
Not explicitly mentioned.
4：Experimental Procedures and Operational Workflow:
The MLP models were trained on input features to predict the best modulation format, evaluated based on classification accuracy and training time.
5：Data Analysis Methods:
Performance was evaluated using accuracy measure, with models chosen based on convergence on validation data.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
