Automatic Modulation Classification (AMC) is a well-known problem in the Radio Frequency (RF) domain. Solving this problem requires determining the modulation of an RF signal. Once the modulation is determined, the signal could be demodulated making it possible to analyse the signal for various purposes. Deep Neural Networks (DNNs) have recently proven to be successful in solving this problem efficiently. However, since deep networks consist of several layers resulting in a high number of trainable parameters, the hardware implementations of these solutions are resource-demanding. In order to address this challenge, we propose an efficient deep neural network referred to as RFNet to tackle the AMC problem efficiently. This network introduces the novel Multiscale Convolutional (MSC) layer to extract robust features in different resolutions. In addition, the network takes advantage of several Separable Convolution Blocks (SCB). These blocks employ pointwise and depth-wise convolutions to reduce network complexity. We further introduce RFNet+ and RFNet++ as extensions of RFNet with fewer number of parameters. These variants include fewer floating-point operations and hence a lower hardware implementation cost. Experimental results using the challenging RadioML 2018 dataset show that RFNet-32++ achieves an average classification accuracy of 56.09% over all Signal-to-Noise Ratios (SNRs) and an accuracy of 92.21% in+20dB SNR using only 3.1K parameters. The small number of parameters makes the RFNet family a promising solution for future AMC systems.
{"title":"RFNet: Fast and efficient neural network for modulation classification of radio frequency signals","authors":"Mohammad Chegini, Pouya Shiri, Amirali Baniasadi","doi":"10.52953/xbpt2357","DOIUrl":"https://doi.org/10.52953/xbpt2357","url":null,"abstract":"Automatic Modulation Classification (AMC) is a well-known problem in the Radio Frequency (RF) domain. Solving this problem requires determining the modulation of an RF signal. Once the modulation is determined, the signal could be demodulated making it possible to analyse the signal for various purposes. Deep Neural Networks (DNNs) have recently proven to be successful in solving this problem efficiently. However, since deep networks consist of several layers resulting in a high number of trainable parameters, the hardware implementations of these solutions are resource-demanding. In order to address this challenge, we propose an efficient deep neural network referred to as RFNet to tackle the AMC problem efficiently. This network introduces the novel Multiscale Convolutional (MSC) layer to extract robust features in different resolutions. In addition, the network takes advantage of several Separable Convolution Blocks (SCB). These blocks employ pointwise and depth-wise convolutions to reduce network complexity. We further introduce RFNet+ and RFNet++ as extensions of RFNet with fewer number of parameters. These variants include fewer floating-point operations and hence a lower hardware implementation cost. Experimental results using the challenging RadioML 2018 dataset show that RFNet-32++ achieves an average classification accuracy of 56.09% over all Signal-to-Noise Ratios (SNRs) and an accuracy of 92.21% in+20dB SNR using only 3.1K parameters. The small number of parameters makes the RFNet family a promising solution for future AMC systems.","PeriodicalId":274720,"journal":{"name":"ITU Journal on Future and Evolving Technologies","volume":"114 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133458740","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jose Rosa, Daniel Granhao, Guilherme Carvalho, Tiago Gon�alves, Monica Figueiredo, Luis Conde Bento, Nuno Paulino, Luis M. Pessoa
Deep learning methods have been shown to be competitive solutions for modulation classification tasks, but suffer from being computationally expensive, limiting their use on embedded devices. We propose a new deep neural network architecture which employs known structures, depth-wise separable convolution and residual connections, as well as a compression methodology, which combined lead to a tiny and fast algorithm for modulation classification. Our compressed model won the first place in ITU's AI/ML in 5G Challenge 2021, achieving 61.73� compression over the challenge baseline and being over 2.6� better than the second best submission. The source code of this work is publicly available at github.com/ITU-AI- ML-in-5G-Challenge/ITU-ML5G-PS-007-BacalhauNet.
{"title":"BacalhauNet: A tiny CNN for lightning-fast modulation classification","authors":"Jose Rosa, Daniel Granhao, Guilherme Carvalho, Tiago Gon�alves, Monica Figueiredo, Luis Conde Bento, Nuno Paulino, Luis M. Pessoa","doi":"10.52953/fywt4006","DOIUrl":"https://doi.org/10.52953/fywt4006","url":null,"abstract":"Deep learning methods have been shown to be competitive solutions for modulation classification tasks, but suffer from being computationally expensive, limiting their use on embedded devices. We propose a new deep neural network architecture which employs known structures, depth-wise separable convolution and residual connections, as well as a compression methodology, which combined lead to a tiny and fast algorithm for modulation classification. Our compressed model won the first place in ITU's AI/ML in 5G Challenge 2021, achieving 61.73� compression over the challenge baseline and being over 2.6� better than the second best submission. The source code of this work is publicly available at github.com/ITU-AI- ML-in-5G-Challenge/ITU-ML5G-PS-007-BacalhauNet.","PeriodicalId":274720,"journal":{"name":"ITU Journal on Future and Evolving Technologies","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125055284","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Guda Blessed, Ibrahim Aliyu, James Agajo, Thiago Lima Sarmento, Cleverson Veloso Nahum, Lucas Novoa, Rebecca Aben-Athar, Mariano Moura, Lucas Matni, Aldebaro Klautau, Deena Mukundan, Divyani R Achari, Mehmet Karaca, Doruk Tayli, �zge Simay Demirci, V. Udaya Sankar, Sai Jnaneswar Juvvisetty, V.M.V.S. Aditya, Abhishek Dandekar, Shabnam Sultana, Jinsul Kim, Vishnu Ram OV
The telecommunication system being a critical pillar of emergency management, intelligent deployment and management of slices in an affected area will help emergency responders. Techniques such as automated management of Machine Learning (ML) pipelines across the edge and emergency responder devices, usage of hierarchical closed-loops, and offloading inference tasks closer to the edge can minimize latencies for first responders in case of emergencies. This study describes the major results from building a Proof of Concept (PoC) for network resource allocation for emergency management using a hierarchical autonomous Artificial Intelligence (AI)/ML-based closed-loops in the mobile network, organized by the Internal Telecommunication Union Focus Group on Autonomous Networks (ITU FG-AN). The background scenario for this PoC included the interaction between a higher closed-loop in the Operations Support System (OSS) and a lower closed-loop in Radio Access Network (RAN) to intelligently share RAN resources between the public and the emergency responder slice. Representation of closed-loop "controllers" in a declarative fashion (intent), triggering "imperative actions" in the "underlay" based on the intent, setup of a data pipeline between various components, and methods of "influencing" lower layer loops using specific logic/models, were some of the essential aspects investigated by various teams. The main conclusions are summarised in this paper, including the significant observations and limitations from the PoC as well as future directions.
{"title":"Network resource allocation for emergency management based on closed-loop analysis","authors":"Guda Blessed, Ibrahim Aliyu, James Agajo, Thiago Lima Sarmento, Cleverson Veloso Nahum, Lucas Novoa, Rebecca Aben-Athar, Mariano Moura, Lucas Matni, Aldebaro Klautau, Deena Mukundan, Divyani R Achari, Mehmet Karaca, Doruk Tayli, �zge Simay Demirci, V. Udaya Sankar, Sai Jnaneswar Juvvisetty, V.M.V.S. Aditya, Abhishek Dandekar, Shabnam Sultana, Jinsul Kim, Vishnu Ram OV","doi":"10.52953/hvpi8935","DOIUrl":"https://doi.org/10.52953/hvpi8935","url":null,"abstract":"The telecommunication system being a critical pillar of emergency management, intelligent deployment and management of slices in an affected area will help emergency responders. Techniques such as automated management of Machine Learning (ML) pipelines across the edge and emergency responder devices, usage of hierarchical closed-loops, and offloading inference tasks closer to the edge can minimize latencies for first responders in case of emergencies. This study describes the major results from building a Proof of Concept (PoC) for network resource allocation for emergency management using a hierarchical autonomous Artificial Intelligence (AI)/ML-based closed-loops in the mobile network, organized by the Internal Telecommunication Union Focus Group on Autonomous Networks (ITU FG-AN). The background scenario for this PoC included the interaction between a higher closed-loop in the Operations Support System (OSS) and a lower closed-loop in Radio Access Network (RAN) to intelligently share RAN resources between the public and the emergency responder slice. Representation of closed-loop \"controllers\" in a declarative fashion (intent), triggering \"imperative actions\" in the \"underlay\" based on the intent, setup of a data pipeline between various components, and methods of \"influencing\" lower layer loops using specific logic/models, were some of the essential aspects investigated by various teams. The main conclusions are summarised in this paper, including the significant observations and limitations from the PoC as well as future directions.","PeriodicalId":274720,"journal":{"name":"ITU Journal on Future and Evolving Technologies","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124535766","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Convolution Neural Network (CNN)-based deep learning models have recently been employed in Automated Modulation Classification (AMC) systems, with excellent results. However, hardware deployment of these CNN-based AMC models is very difficult due to their large size, floating point weights and activations, and real-time processing requirements in hardware such as Field Programmable Gate Arrays (FPGAs). In this study, we designed CNN-based AMC techniques for complex-valued temporal radio signal domains and made them less complex with a small memory footprint for FPGA implementation. This work mainly focuses on quantized CNN, low precision mathematics, and quantization-aware CNN training to overcome the problem of larger model sizes, floating-point weights, and activations. Low precision weights, activations, and quantized CNN, on the other hand, have a considerable impact on the accuracy of the model. Thus, we propose an iterative pruning-based training mechanism to maintain the overall accuracy above a certain threshold while decreasing the model size for hardware implementation. The proposed schemes are 21.55 times less complex and achieve at least 1.6% higher accuracy than the baseline. Moreover, results show that our convolution layer-based Quantized Modulation Classification Network (QMCNet) with pruning has 92.01% less multiply-accumulate bit operations (bit_operations), 61.39% less activation bits, and 87.58% less weight bits than the 8 bit quantized baseline model whereas the quantized and pruned Residual-Unit based model (RUNet) has 95.36% less bit_operations, 29.97% less activation bits and 98.22% less weight bits than the 8 bit quantized baseline model.
{"title":"Low-precision deep-learning-based automatic modulation recognition system","authors":"Satish Kumar, Aakash Agarwal, Neeraj Varshney, Rajarshi Mahapatra","doi":"10.52953/ctyj2699","DOIUrl":"https://doi.org/10.52953/ctyj2699","url":null,"abstract":"Convolution Neural Network (CNN)-based deep learning models have recently been employed in Automated Modulation Classification (AMC) systems, with excellent results. However, hardware deployment of these CNN-based AMC models is very difficult due to their large size, floating point weights and activations, and real-time processing requirements in hardware such as Field Programmable Gate Arrays (FPGAs). In this study, we designed CNN-based AMC techniques for complex-valued temporal radio signal domains and made them less complex with a small memory footprint for FPGA implementation. This work mainly focuses on quantized CNN, low precision mathematics, and quantization-aware CNN training to overcome the problem of larger model sizes, floating-point weights, and activations. Low precision weights, activations, and quantized CNN, on the other hand, have a considerable impact on the accuracy of the model. Thus, we propose an iterative pruning-based training mechanism to maintain the overall accuracy above a certain threshold while decreasing the model size for hardware implementation. The proposed schemes are 21.55 times less complex and achieve at least 1.6% higher accuracy than the baseline. Moreover, results show that our convolution layer-based Quantized Modulation Classification Network (QMCNet) with pruning has 92.01% less multiply-accumulate bit operations (bit_operations), 61.39% less activation bits, and 87.58% less weight bits than the 8 bit quantized baseline model whereas the quantized and pruned Residual-Unit based model (RUNet) has 95.36% less bit_operations, 29.97% less activation bits and 98.22% less weight bits than the 8 bit quantized baseline model.","PeriodicalId":274720,"journal":{"name":"ITU Journal on Future and Evolving Technologies","volume":"197 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121879982","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Junior Momo Ziazet, Charles Boudreau, Brigitte Jaumard, Huy Duong
With recent advances in the field of Machine Learning (ML), a multitude of problems related to communication systems and networks can be solved with data-driven solutions. Since data in these systems is mostly represented as graphs, Graph-based Neural Networks (GNNs) are a good candidate for solving such problems. These GNNs can be used as a computer network modeling technique to build models that accurately estimate the Key Performance Indicators (KPI) such as delay or jitter in real network scenarios in order to ensure their requirements in terms of service assurance. To build GNN solutions with higher accuracy, low computational resource requirements, and easy deployment of synthetic network training results into real-world networks, it is more than necessary to develop efficient and effective GNN models. This paper presents a GNN model capable of accurately estimating the average delay per flow in networks. By designing scale-independent features and using notions from queuing theory, the proposed model successfully generalizes to large size topologies, routing configurations, and traffic matrices not seen during the training phase.
{"title":"Addressing RouteNet scalability through input and output design","authors":"Junior Momo Ziazet, Charles Boudreau, Brigitte Jaumard, Huy Duong","doi":"10.52953/giod4389","DOIUrl":"https://doi.org/10.52953/giod4389","url":null,"abstract":"With recent advances in the field of Machine Learning (ML), a multitude of problems related to communication systems and networks can be solved with data-driven solutions. Since data in these systems is mostly represented as graphs, Graph-based Neural Networks (GNNs) are a good candidate for solving such problems. These GNNs can be used as a computer network modeling technique to build models that accurately estimate the Key Performance Indicators (KPI) such as delay or jitter in real network scenarios in order to ensure their requirements in terms of service assurance. To build GNN solutions with higher accuracy, low computational resource requirements, and easy deployment of synthetic network training results into real-world networks, it is more than necessary to develop efficient and effective GNN models. This paper presents a GNN model capable of accurately estimating the average delay per flow in networks. By designing scale-independent features and using notions from queuing theory, the proposed model successfully generalizes to large size topologies, routing configurations, and traffic matrices not seen during the training phase.","PeriodicalId":274720,"journal":{"name":"ITU Journal on Future and Evolving Technologies","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134123799","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In recent years, indoor localization using wireless systems has been an important area of research for its applications towards health, security and the tracking of users. A Global Positioning System (GPS) is considered as the best solution for localization for outdoor scenarios but it fails to provide accurate positioning for indoor scenarios. Wi-Fi fingerprinting methods using received signal strength from multiple access points are popular for solving indoor localization problem. As the wireless systems move towards higher frequencies, higher bandwidth and a large antenna array, sensing has also become feasible along with communication, which is an important research area towards 6G named as Integrated Communication And Sensing (ISAC). ISAC relies on sensing parameter estimations, such as estimation of fine range, Doppler and angular information which contains the signature of the surrounding objects. A localization problem can be solved by analysing the sensing parameters. In this paper, we propose a solution for the localization problem for IEEE 802.11ay WLAN systems based on signal processing and Machine Learning (ML) in indoor scenarios. (...)
{"title":"AI-based indoor localization using mmWave MIMO channel at 60 GHz","authors":"Shubham Khunteta, Ashok Kumar Reddy, Avani Agrawal","doi":"10.52953/aorf8087","DOIUrl":"https://doi.org/10.52953/aorf8087","url":null,"abstract":"In recent years, indoor localization using wireless systems has been an important area of research for its applications towards health, security and the tracking of users. A Global Positioning System (GPS) is considered as the best solution for localization for outdoor scenarios but it fails to provide accurate positioning for indoor scenarios. Wi-Fi fingerprinting methods using received signal strength from multiple access points are popular for solving indoor localization problem. As the wireless systems move towards higher frequencies, higher bandwidth and a large antenna array, sensing has also become feasible along with communication, which is an important research area towards 6G named as Integrated Communication And Sensing (ISAC). ISAC relies on sensing parameter estimations, such as estimation of fine range, Doppler and angular information which contains the signature of the surrounding objects. A localization problem can be solved by analysing the sensing parameters. In this paper, we propose a solution for the localization problem for IEEE 802.11ay WLAN systems based on signal processing and Machine Learning (ML) in indoor scenarios. (...)","PeriodicalId":274720,"journal":{"name":"ITU Journal on Future and Evolving Technologies","volume":"133 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130797047","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Priyanshu M, Venkatesh Subramanya Iyer Giri, Shachi P, Geetishree Mishra, Suma M N
Radio link sustainability gets affected by weather adversities such as snow, fog, cloud, rain, thunderstorm, etc. A proactive solution in radio link failure scenarios is necessary to overcome economic loss and maintain the Quality of Service (QoS). To address the issue, our work contributes towards building a machine-learning-based solution to predict the radio link failure when generic regional weather forecast data, key performance indices of radio link and spatial nature of the data are available. After rigorous data preprocessing, ensembling models like logistic regression, random forest, light BGM, XGBoost and gradient boosting classifiers were trained to predict the Radio Link Failure (RLF) for two cases i.e., day-1-predict and day-5-predict. Since it is a classification use case, the metrics used for our work are precision, recall, and F1 score. The performance of the gradient boosting classifier was better as compared to the other models with an F1 score of 0.95 for both day-1-predict and day-5-predict.
{"title":"AI powered solution for radio link failure prediction based on link features and weather forecast","authors":"Priyanshu M, Venkatesh Subramanya Iyer Giri, Shachi P, Geetishree Mishra, Suma M N","doi":"10.52953/odqq8049","DOIUrl":"https://doi.org/10.52953/odqq8049","url":null,"abstract":"Radio link sustainability gets affected by weather adversities such as snow, fog, cloud, rain, thunderstorm, etc. A proactive solution in radio link failure scenarios is necessary to overcome economic loss and maintain the Quality of Service (QoS). To address the issue, our work contributes towards building \u0000a machine-learning-based solution to predict the radio link failure when generic regional weather forecast data, key performance indices of radio link and spatial nature of the data are available. After rigorous data preprocessing, ensembling models like logistic regression, random forest, light BGM, XGBoost and gradient boosting classifiers were trained to predict the Radio Link Failure (RLF) for two cases i.e., day-1-predict and day-5-predict. Since it is a classification use case, the metrics used for our work are precision, recall, and F1 score. The performance of the gradient boosting classifier was better as compared to the other models with an F1 score of 0.95 for both day-1-predict and day-5-predict.","PeriodicalId":274720,"journal":{"name":"ITU Journal on Future and Evolving Technologies","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125691326","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper shows a flexible orchestration solution for deploying Cooperative, Connected, and Automated Mobility (CCAM) services in 5G and beyond networks. This solution is based on the concepts of federation and hierarchy of orchestration functions. The federated approach is leveraged to cope with the differentiated complexity operation when multiple network operators are considered, whereas the hierarchical approach addresses the issue of jointly orchestrating multiple edge platforms in the network of a single operator. In this complex orchestration architecture, the main contribution of this paper consists of the design and implementation of an Abstraction and Adaptation Layer (AAL) for edge clouds, a new component enabling a truly cooperative and coordinated orchestration between different edge systems, characterized by appreciable experimental performance in terms of latency.
{"title":"An edge abstraction layer enabling federated and hierarchical orchestration of CCAM services in 5G and beyond networks","authors":"Mauro Femminella, Gianluca Reali","doi":"10.52953/lnav1342","DOIUrl":"https://doi.org/10.52953/lnav1342","url":null,"abstract":"This paper shows a flexible orchestration solution for deploying Cooperative, Connected, and Automated Mobility (CCAM) services in 5G and beyond networks. This solution is based on the concepts of federation and hierarchy of orchestration functions. The federated approach is leveraged to cope with the differentiated complexity operation when multiple network operators are considered, whereas the hierarchical approach addresses the issue of jointly orchestrating multiple edge platforms in the network of a single operator. In this complex orchestration architecture, the main contribution of this paper consists of the design and implementation of an Abstraction and Adaptation Layer (AAL) for edge clouds, a new component enabling a truly cooperative and coordinated orchestration between different edge systems, characterized by appreciable experimental performance in terms of latency.","PeriodicalId":274720,"journal":{"name":"ITU Journal on Future and Evolving Technologies","volume":"151 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132231820","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Nethraa Sivakumar, Pooja Srinivasan, Nikhil Viswanath, Venkateswaran N
Stable and high-quality Internet connectivity is mandatory for 5G mobile networks. Network disruption may occur due to unexpected variations in environmental conditions such as weather, wind, and natural or man-made surroundings, and the influence of the defect is quite severe. Prediction of such undesirable events at a low cost can boost 5G communication reliability, massive network capacity, and decreased latency. This research work makes use of novel preprocessing and feature engineering techniques, followed by a trained decision tree model to predict the occurrence of Radio Link Failure (RLF). This system is designed to predict RLF for not just the next day, but also any of the next 5 days. This prediction supports reliance and increasing demand for good Internet connectivity. In order to achieve accurate RLF prediction, comprehensive data has been used which undergoes preprocessing. To account for the influence of surrounding weather conditions on radio links, the proposed system makes use of information from the past i.e., previous RLFs, and the information from the future i.e., the weather forecast from the weather station around the radio link station. The decision tree model was trained with the integration of feature engineering. A macro-averaged F1-score of 70% and 77% were obtained for RLF prediction for the next day and RLF prediction for the next 5 days, respectively. The results show improvement in performance after the incorporation of feature engineering in the pipeline. Further, an additional metric termed G-Mean is introduced in the paper. Owing to the high imbalance in the dataset, this metric was found to provide a more realistic representation of the results. The G-Mean score was found to be 98.69% and 92.89% for RLF prediction for the next day and RLF prediction for the next 5 days, respectively.
{"title":"Decision tree-based radio link failure prediction for 5G communication reliability","authors":"Nethraa Sivakumar, Pooja Srinivasan, Nikhil Viswanath, Venkateswaran N","doi":"10.52953/lzlj8762","DOIUrl":"https://doi.org/10.52953/lzlj8762","url":null,"abstract":"Stable and high-quality Internet connectivity is mandatory for 5G mobile networks. Network disruption may occur due to unexpected variations in environmental conditions such as weather, wind, and natural or man-made surroundings, and the influence of the defect is quite severe. Prediction of such undesirable events at a low cost can boost 5G communication reliability, massive network capacity, and decreased latency. This research work makes use of novel preprocessing and feature engineering techniques, followed by a trained decision tree model to predict the occurrence of Radio Link Failure (RLF). This system is designed to predict RLF for not just the next day, but also any of the next 5 days. This prediction supports reliance and increasing demand for good Internet connectivity. In order to achieve accurate RLF prediction, comprehensive data has been used which undergoes preprocessing. To account for the influence of surrounding weather conditions on radio links, the proposed system makes use of information from the past i.e., previous RLFs, and the information from the future i.e., the weather forecast from the weather station around the radio link station. The decision tree model was trained with the integration of feature engineering. A macro-averaged F1-score of 70% and 77% were obtained for RLF prediction for the next day and RLF prediction for the next 5 days, respectively. The results show improvement in performance after the incorporation of feature engineering in the pipeline. Further, an additional metric termed G-Mean is introduced in the paper. Owing to the high imbalance in the dataset, this metric was found to provide a more realistic representation of the results. The G-Mean score was found to be 98.69% and 92.89% for RLF prediction for the next day and RLF prediction for the next 5 days, respectively.","PeriodicalId":274720,"journal":{"name":"ITU Journal on Future and Evolving Technologies","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128548935","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In the past few years, there have been numerous demonstrations of neural networks outperforming traditional signal processing methods in communications, notably for Automatic Modulation Classification (AMC). Despite the increase in accuracy, these algorithms are notoriously infeasible for integrating into edge computing applications. In this work, we propose an enhanced version of a simple neural network pruning technique, Iterative Magnitude Pruning (IMP), called Feedback Magnitude Pruning (FMP) and demonstrate its effectiveness for the "Lightning-Fast Modulation Classification with Hardware-Effficient Neural Network" 2021 AI for Good: Machine Learning in 5G Challenge hosted by the International Telecommunications Union (ITU) and Xilinx. IMP achieved a compression ratio of 9.313, while our proposed FMP achieved a compression ratio of 831 and normalized cost of 0.0419. Our FMP result was awarded second place, demonstrating the compression and classification accuracy benefits of pruning with feedback.
在过去的几年中，神经网络在通信领域的表现优于传统的信号处理方法，特别是在自动调制分类(AMC)方面。尽管准确性有所提高，但众所周知，这些算法在集成到边缘计算应用程序中是不可行的。在这项工作中，我们提出了一种简单的神经网络修剪技术的增强版本，迭代幅度修剪(IMP)，称为反馈幅度修剪(FMP)，并证明了其在国际电信联盟(ITU)和赛灵思主办的“硬件高效神经网络闪电般的快速调制分类”2021年AI for Good:机器学习5G挑战中的有效性。IMP实现了9.313的压缩比，而我们提出的FMP实现了831的压缩比和0.0419的归一化成本。我们的FMP结果获得了第二名，证明了带有反馈的修剪在压缩和分类精度方面的好处。
{"title":"Neural network compression with feedback magnitude pruning for automatic modulation classification","authors":"Jakob Krzyston, Rajib Bhattacharjea, Andrew Stark","doi":"10.52953/eujf4214","DOIUrl":"https://doi.org/10.52953/eujf4214","url":null,"abstract":"In the past few years, there have been numerous demonstrations of neural networks outperforming traditional signal processing methods in communications, notably for Automatic Modulation Classification (AMC). Despite the increase in accuracy, these algorithms are notoriously infeasible for integrating into edge computing applications. In this work, we propose an enhanced version of a simple neural network pruning technique, Iterative Magnitude Pruning (IMP), called Feedback Magnitude Pruning (FMP) and demonstrate its effectiveness for the \"Lightning-Fast Modulation Classification with Hardware-Effficient Neural Network\" 2021 AI for Good: Machine Learning in 5G Challenge hosted by the International Telecommunications Union (ITU) and Xilinx. IMP achieved a compression ratio of 9.313, while our proposed FMP achieved a compression ratio of 831 and normalized cost of 0.0419. Our FMP result was awarded second place, demonstrating the compression and classification accuracy benefits of pruning with feedback.","PeriodicalId":274720,"journal":{"name":"ITU Journal on Future and Evolving Technologies","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123813357","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：