Pub Date : 2022-05-24DOI: 10.1109/NCC55593.2022.9806759
V. S. Bhadouria, Ritesh Kumar, Monika Aggarwal
Characterizing the UWA channel is critical for designing a robust communication receiver. Due to the signif-icant delay spread, underwater channels make communication difficult. The time-reversal mirror improves the channel char-acteristics by reducing delay spread and increasing coherence bandwidth. This paper analyses and quantifies a time reversal mirror (TRM) effect on an underwater acoustic channel. The delay spread decreases as the number of receivers increases, but this decrease is asymptotic, meaning that regardless of the receiver geometry, the delay spread converges to a fixed non-zero value. Additionally, this analysis establishes that the TRM effectiveness is dependent on the water column depth and the distance between the transmitter and receiver. As the number of receivers increases, the effectiveness of TRM approaches the same value regardless of the water column depth and the distance between the transmitter and receiver. Additionally, the effect of TRM on the spread of delays is validated in the actual sea environment.
{"title":"On study and analysis of the impact of the time reversal mirror on characteristics of the underwater acoustic channel","authors":"V. S. Bhadouria, Ritesh Kumar, Monika Aggarwal","doi":"10.1109/NCC55593.2022.9806759","DOIUrl":"https://doi.org/10.1109/NCC55593.2022.9806759","url":null,"abstract":"Characterizing the UWA channel is critical for designing a robust communication receiver. Due to the signif-icant delay spread, underwater channels make communication difficult. The time-reversal mirror improves the channel char-acteristics by reducing delay spread and increasing coherence bandwidth. This paper analyses and quantifies a time reversal mirror (TRM) effect on an underwater acoustic channel. The delay spread decreases as the number of receivers increases, but this decrease is asymptotic, meaning that regardless of the receiver geometry, the delay spread converges to a fixed non-zero value. Additionally, this analysis establishes that the TRM effectiveness is dependent on the water column depth and the distance between the transmitter and receiver. As the number of receivers increases, the effectiveness of TRM approaches the same value regardless of the water column depth and the distance between the transmitter and receiver. Additionally, the effect of TRM on the spread of delays is validated in the actual sea environment.","PeriodicalId":403870,"journal":{"name":"2022 National Conference on Communications (NCC)","volume":"66 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123829888","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}
Pub Date : 2022-05-24DOI: 10.1109/NCC55593.2022.9806732
Abinay Reddy Naini, Achuth Rao M V, P. Ghosh
In this work, we propose an iterative optimization algorithm to learn a feature mapping (FM) from the whispered to neutral speech features. Such an FM can be used to improve the performance of speaker verification (SV) systems when presented with a whispered speech. In one of previous works, the equal error rate (EER) in an SV task has been shown to improve by ~24%. based on an FM network trained using a cosine similarity based loss function over that using a mean squared error based objective function. As the mapped whispered features obtained in this manner may not lie in the trained i-vector space, we, in this work, iteratively optimize the i-vector space likelihood (by updating T-matrix) and a cosine similarity based loss function for learning the parameters of the FM network. The proposed iterative optimization improves the EER by ~26% compared to when the FM network parameters are learned based on only cosine similarity based loss function without any T-matrix update, which is a special case of the proposed iterative optimization.
{"title":"Whisper to Neutral Mapping Using I-Vector Space Likelihood and a Cosine Similarity Based Iterative Optimization for Whispered Speaker Verification","authors":"Abinay Reddy Naini, Achuth Rao M V, P. Ghosh","doi":"10.1109/NCC55593.2022.9806732","DOIUrl":"https://doi.org/10.1109/NCC55593.2022.9806732","url":null,"abstract":"In this work, we propose an iterative optimization algorithm to learn a feature mapping (FM) from the whispered to neutral speech features. Such an FM can be used to improve the performance of speaker verification (SV) systems when presented with a whispered speech. In one of previous works, the equal error rate (EER) in an SV task has been shown to improve by ~24%. based on an FM network trained using a cosine similarity based loss function over that using a mean squared error based objective function. As the mapped whispered features obtained in this manner may not lie in the trained i-vector space, we, in this work, iteratively optimize the i-vector space likelihood (by updating T-matrix) and a cosine similarity based loss function for learning the parameters of the FM network. The proposed iterative optimization improves the EER by ~26% compared to when the FM network parameters are learned based on only cosine similarity based loss function without any T-matrix update, which is a special case of the proposed iterative optimization.","PeriodicalId":403870,"journal":{"name":"2022 National Conference on Communications (NCC)","volume":"108 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134497518","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}
Pub Date : 2022-05-24DOI: 10.1109/NCC55593.2022.9806766
Shaik John Saida, S. Ari
Ocean eddies are a common occurrence in ocean water circulation. They have an enormous impact on the marine ecosystem. One of the most active study topics in physical oceanography is ocean eddy detection. Although using deep learning algorithms to detect eddies is a recent trend, it is still in its infancy. In this paper, an attention mechanism-based ocean eddy detection approach using deep learning is proposed. Attention mechanism has spatial and channel attention modules that are cascaded to convolution blocks-based encoder model to simulate spatial and channel semantic interdependencies. In the spatial attention module, the feature at each point is aggregated selectively by the sum of the features at all positions. The channel attention module aggregates related data from all channel maps to selectively highlight interdependent channel maps. The original feature map and the feature map obtained through the attention mechanism are appended to enhance the feature representation further, resulting in more accurate segmentation results. The findings of the experiments show that adopting an attention-based deep framework improves eddy recognition accuracy significantly.
{"title":"Automatic Detection of Ocean Eddy based on Deep Learning Technique with Attention Mechanism","authors":"Shaik John Saida, S. Ari","doi":"10.1109/NCC55593.2022.9806766","DOIUrl":"https://doi.org/10.1109/NCC55593.2022.9806766","url":null,"abstract":"Ocean eddies are a common occurrence in ocean water circulation. They have an enormous impact on the marine ecosystem. One of the most active study topics in physical oceanography is ocean eddy detection. Although using deep learning algorithms to detect eddies is a recent trend, it is still in its infancy. In this paper, an attention mechanism-based ocean eddy detection approach using deep learning is proposed. Attention mechanism has spatial and channel attention modules that are cascaded to convolution blocks-based encoder model to simulate spatial and channel semantic interdependencies. In the spatial attention module, the feature at each point is aggregated selectively by the sum of the features at all positions. The channel attention module aggregates related data from all channel maps to selectively highlight interdependent channel maps. The original feature map and the feature map obtained through the attention mechanism are appended to enhance the feature representation further, resulting in more accurate segmentation results. The findings of the experiments show that adopting an attention-based deep framework improves eddy recognition accuracy significantly.","PeriodicalId":403870,"journal":{"name":"2022 National Conference on Communications (NCC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130170346","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}
Semantic segmentation is a crucial algorithm for identifying various objects in the surrounding of an autonomous vehicle. However, due to the limited size of real-world datasets, domain adaptation is employed. Hence, the models are made to adapt to real-world settings while being trained on large-scale synthetic datasets. In domain adaptation, domain-invariant features play a significant role in learning domain agnostic representations for each predefined category. While most of the prior work focuses on decreasing the distance between the domains, the works that utilize contrastive objectives for learning domain-invariant features depend heavily on the augmentations used. In this work, we completely eradicate the requirement of explicit data augmentations. We hypothesize that real-world images and their corresponding synthetic images are different views of the same abstract representation. To enhance the quality of domain-invariant features, we increase the mutual information between the two inputs. We first validate our hypothesis on the classification task using the standard datasets; Office31 and VisDA-2017. Further, we perform quantitative and qualitative analysis on the segmentation task using SYNTHIA, GTA and Cityscapes datasets.
{"title":"Contrastive Learning-Based Domain Adaptation for Semantic Segmentation","authors":"Rishika Bhagwatkar, Saurabh Kemekar, Vinay Domatoti, Khursheed Munir Khan, Anamika Singh","doi":"10.1109/NCC55593.2022.9806740","DOIUrl":"https://doi.org/10.1109/NCC55593.2022.9806740","url":null,"abstract":"Semantic segmentation is a crucial algorithm for identifying various objects in the surrounding of an autonomous vehicle. However, due to the limited size of real-world datasets, domain adaptation is employed. Hence, the models are made to adapt to real-world settings while being trained on large-scale synthetic datasets. In domain adaptation, domain-invariant features play a significant role in learning domain agnostic representations for each predefined category. While most of the prior work focuses on decreasing the distance between the domains, the works that utilize contrastive objectives for learning domain-invariant features depend heavily on the augmentations used. In this work, we completely eradicate the requirement of explicit data augmentations. We hypothesize that real-world images and their corresponding synthetic images are different views of the same abstract representation. To enhance the quality of domain-invariant features, we increase the mutual information between the two inputs. We first validate our hypothesis on the classification task using the standard datasets; Office31 and VisDA-2017. Further, we perform quantitative and qualitative analysis on the segmentation task using SYNTHIA, GTA and Cityscapes datasets.","PeriodicalId":403870,"journal":{"name":"2022 National Conference on Communications (NCC)","volume":"222 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115785452","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}
Pub Date : 2022-05-24DOI: 10.1109/NCC55593.2022.9806797
Amala Sonny, Abhinav Kumar
Wi-Fi based indoor localization has gained much attention around the globe due to its widespread reach and availability. Amongst several possible approaches using Wi-Fi signals, fingerprint image-based approach has become popular due to its low hardware requirements. Further, this approach can be used alone or along with other positioning systems for indoor localization. However, a multi-building, multi-floor indoor positioning system with high localization accuracy is required. Motivated by this, we propose a Convolutional Neural Networks (CNN)-based approach. For feature extraction and classification, a multi-output multi-label sequential 2D-CNN classifier is developed and implemented. The system is able to predict the location of the user by combining the classification output from the multi-output model. This approach is verified on the publicly available UJIIndoorLoc database. The system offers an average accuracy of 97% in indoor localization.
{"title":"Fingerprint Image-Based Multi-Building 3D Indoor Wi-Fi Localization Using Convolutional Neural Networks","authors":"Amala Sonny, Abhinav Kumar","doi":"10.1109/NCC55593.2022.9806797","DOIUrl":"https://doi.org/10.1109/NCC55593.2022.9806797","url":null,"abstract":"Wi-Fi based indoor localization has gained much attention around the globe due to its widespread reach and availability. Amongst several possible approaches using Wi-Fi signals, fingerprint image-based approach has become popular due to its low hardware requirements. Further, this approach can be used alone or along with other positioning systems for indoor localization. However, a multi-building, multi-floor indoor positioning system with high localization accuracy is required. Motivated by this, we propose a Convolutional Neural Networks (CNN)-based approach. For feature extraction and classification, a multi-output multi-label sequential 2D-CNN classifier is developed and implemented. The system is able to predict the location of the user by combining the classification output from the multi-output model. This approach is verified on the publicly available UJIIndoorLoc database. The system offers an average accuracy of 97% in indoor localization.","PeriodicalId":403870,"journal":{"name":"2022 National Conference on Communications (NCC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131007942","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}
Pub Date : 2022-05-24DOI: 10.1109/NCC55593.2022.9806802
Shrikant Sharma, A. Girish, Nikhar P. Rakhashia, V. Gadre, Shaan ul Haque, Aseer Ansari, R. B. Pachori, P. Radhakrishna, Peeyush Sahay
In this paper, we perform a theoretical analysis of an inverse Radon transform-based micro-Doppler parameter es-timation algorithm. For a multicomponent micro-Doppler signal, no mathematical expression was proposed in this algorithm to find the number of frequency terms to be dropped for efficient elimination of estimated micro-Doppler components. Hence, we first derive an expression for the number of frequency terms to set to zero for efficient elimination of estimated micro-Doppler components by exploiting standard Bessel function properties. We verify our result through simulations with up to three targets, even in the presence of noise. We also provide an analysis of the limiting performance of the algorithm for two targets as the parameters are made close to each other.
{"title":"Theoretical Analysis of an Inverse Radon Transform Based Multicomponent Micro-Doppler Parameter Estimation Algorithm","authors":"Shrikant Sharma, A. Girish, Nikhar P. Rakhashia, V. Gadre, Shaan ul Haque, Aseer Ansari, R. B. Pachori, P. Radhakrishna, Peeyush Sahay","doi":"10.1109/NCC55593.2022.9806802","DOIUrl":"https://doi.org/10.1109/NCC55593.2022.9806802","url":null,"abstract":"In this paper, we perform a theoretical analysis of an inverse Radon transform-based micro-Doppler parameter es-timation algorithm. For a multicomponent micro-Doppler signal, no mathematical expression was proposed in this algorithm to find the number of frequency terms to be dropped for efficient elimination of estimated micro-Doppler components. Hence, we first derive an expression for the number of frequency terms to set to zero for efficient elimination of estimated micro-Doppler components by exploiting standard Bessel function properties. We verify our result through simulations with up to three targets, even in the presence of noise. We also provide an analysis of the limiting performance of the algorithm for two targets as the parameters are made close to each other.","PeriodicalId":403870,"journal":{"name":"2022 National Conference on Communications (NCC)","volume":"121 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133229106","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}
Pub Date : 2022-05-24DOI: 10.1109/NCC55593.2022.9806729
Rashmi Kamran, P. Jha, Shwetha Kiran, A. Karandikar, P. Chaporkar, Anindya Saha, Arindam Chakraborty
Increased usage of video consumption along with a host of new services such as software download over wireless networks, group communications, and Internet of Things (IoT) applications have created a need for support of Multicast Broadcast Services (MBS) in wireless networks. While the Third Generation Partnership Project (3GPP) is defining its own mechanism for MBS support in Fifth Generation (5G) system, supplementing the native 5G MBS support with non-3GPP Broadcast Networks may bring additional advantages. A unique characteristic of the 3GPP 5G System (5GS) architecture is the existence of a converged core, capable of supporting diverse access technologies, 3GPP and non-3GPP access technologies in a uniform manner. The 5GS also supports multiple integration points for non-3GPP access networks. These may be utilized for its integration with non-3GPP broadcast networks such as non-3GPP satellite access networks and digital terrestrial broadcast networks enabling it to harness them for multicast broadcast service delivery. In this article, we review the upcoming 3GPP 5G MBS standards along with some of its limitations. We also present state of the art standardization initiatives towards convergence of non-3GPP broadcast networks with the 5GS including our proposals submitted to standards organizations.
{"title":"A Survey on Multicast Broadcast Services in 5G and Beyond","authors":"Rashmi Kamran, P. Jha, Shwetha Kiran, A. Karandikar, P. Chaporkar, Anindya Saha, Arindam Chakraborty","doi":"10.1109/NCC55593.2022.9806729","DOIUrl":"https://doi.org/10.1109/NCC55593.2022.9806729","url":null,"abstract":"Increased usage of video consumption along with a host of new services such as software download over wireless networks, group communications, and Internet of Things (IoT) applications have created a need for support of Multicast Broadcast Services (MBS) in wireless networks. While the Third Generation Partnership Project (3GPP) is defining its own mechanism for MBS support in Fifth Generation (5G) system, supplementing the native 5G MBS support with non-3GPP Broadcast Networks may bring additional advantages. A unique characteristic of the 3GPP 5G System (5GS) architecture is the existence of a converged core, capable of supporting diverse access technologies, 3GPP and non-3GPP access technologies in a uniform manner. The 5GS also supports multiple integration points for non-3GPP access networks. These may be utilized for its integration with non-3GPP broadcast networks such as non-3GPP satellite access networks and digital terrestrial broadcast networks enabling it to harness them for multicast broadcast service delivery. In this article, we review the upcoming 3GPP 5G MBS standards along with some of its limitations. We also present state of the art standardization initiatives towards convergence of non-3GPP broadcast networks with the 5GS including our proposals submitted to standards organizations.","PeriodicalId":403870,"journal":{"name":"2022 National Conference on Communications (NCC)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126897613","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}
Pub Date : 2022-05-24DOI: 10.1109/NCC55593.2022.9806773
Pranaya Yellu, Satyam Singh, S. Joshi, R. Sarkar, Soumya Jana
Sarcopenia is increasingly identified as a correlate of frailty and ageing and associated with an increased likelihood of falls, fracture, frailty and mortality. The gold standard for the sarcopenia evaluation in computed tomography (CT) scan was psoas muscle area (PMA) measurement. In this paper, we proposed an automated deep learning approach to find the muscle volume and assessed the correlation between PMA and muscle volume in the chest CT. This alternate muscle volume metric becomes significant since most chest CT scans taken to assess lung diseases might not consist of psoas muscle but consists of other muscles, and it would therefore not be possible to assess sarcopenia in chest CT. Our results show a good correlation between the psoas muscle area and the muscle volume produced over specific anatomical landmarks by segmenting the muscle tissue using the 2D U-Net segmentation model, strengthening our proposition. Along with the muscle volume, we have also found the volume of peripheral fat and have shown there exists a correlation between them which could be helpful for nutritional evaluation.
{"title":"Automated Volumetric Examination of Muscle for Sarcopenia Assessment in CT Scan: Generalization of Psoas-based Approach","authors":"Pranaya Yellu, Satyam Singh, S. Joshi, R. Sarkar, Soumya Jana","doi":"10.1109/NCC55593.2022.9806773","DOIUrl":"https://doi.org/10.1109/NCC55593.2022.9806773","url":null,"abstract":"Sarcopenia is increasingly identified as a correlate of frailty and ageing and associated with an increased likelihood of falls, fracture, frailty and mortality. The gold standard for the sarcopenia evaluation in computed tomography (CT) scan was psoas muscle area (PMA) measurement. In this paper, we proposed an automated deep learning approach to find the muscle volume and assessed the correlation between PMA and muscle volume in the chest CT. This alternate muscle volume metric becomes significant since most chest CT scans taken to assess lung diseases might not consist of psoas muscle but consists of other muscles, and it would therefore not be possible to assess sarcopenia in chest CT. Our results show a good correlation between the psoas muscle area and the muscle volume produced over specific anatomical landmarks by segmenting the muscle tissue using the 2D U-Net segmentation model, strengthening our proposition. Along with the muscle volume, we have also found the volume of peripheral fat and have shown there exists a correlation between them which could be helpful for nutritional evaluation.","PeriodicalId":403870,"journal":{"name":"2022 National Conference on Communications (NCC)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122070754","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}
Pub Date : 2022-05-24DOI: 10.1109/NCC55593.2022.9806469
D. Shivakrishna, V. Lalitha
Product codes are a class of codes which have generator matrices as the tensor product of the component codes and the codeword itself can be represented as an (m × n) array, where the component codes themselves are referred to as the row and column codes. Maximally recoverable product codes (MRPCs) are a class of codes which can recover from all information theoretically recoverable erasure patterns, given the $a$ column and $b$ row constraints imposed by the code. In this work, we derive puncturing and shortening properties of maximally recoverable product codes. We give a sufficient condition to characterize a certain subclass of erasure patterns as correctable and another necessary condition to characterize another subclass of erasure patterns as not correctable. In an earlier work, higher order MDS codes denoted by MDS(l) have been defined in terms of generic matrices and these codes have been shown to be constituent row codes for maximally recoverable product codes for the case of $a$ = 1. We derive a certain inclusion-exclusion type principle for characterizing the dimension of intersection spaces of generic matrices. Applying this, we formally derive a relation between MDS(3) codes and points/lines of the associated projective space.
{"title":"Properties of Maximally Recoverable Product Codes and Higher Order MDS Codes","authors":"D. Shivakrishna, V. Lalitha","doi":"10.1109/NCC55593.2022.9806469","DOIUrl":"https://doi.org/10.1109/NCC55593.2022.9806469","url":null,"abstract":"Product codes are a class of codes which have generator matrices as the tensor product of the component codes and the codeword itself can be represented as an (m × n) array, where the component codes themselves are referred to as the row and column codes. Maximally recoverable product codes (MRPCs) are a class of codes which can recover from all information theoretically recoverable erasure patterns, given the $a$ column and $b$ row constraints imposed by the code. In this work, we derive puncturing and shortening properties of maximally recoverable product codes. We give a sufficient condition to characterize a certain subclass of erasure patterns as correctable and another necessary condition to characterize another subclass of erasure patterns as not correctable. In an earlier work, higher order MDS codes denoted by MDS(l) have been defined in terms of generic matrices and these codes have been shown to be constituent row codes for maximally recoverable product codes for the case of $a$ = 1. We derive a certain inclusion-exclusion type principle for characterizing the dimension of intersection spaces of generic matrices. Applying this, we formally derive a relation between MDS(3) codes and points/lines of the associated projective space.","PeriodicalId":403870,"journal":{"name":"2022 National Conference on Communications (NCC)","volume":"95 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131474586","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}
Pub Date : 2022-05-24DOI: 10.1109/NCC55593.2022.9806778
Sawan Singh Mahara, M. Shruti, B. Bharath
The paper proposes a novel Federated Learning (FL) algorithm involving signed gradient as feedback to reduce communication overhead. The Multi-task nature of the algorithm provides each device a custom neural network after completion. Towards improving the performance, a weighted average loss across devices is proposed which considers the similarity between their data distributions. A Probably Approximately Correct (PAC) bound on the true loss in terms of the proposed empirical loss is derived. The bound is in terms of (i) Rademacher complexity, (ii) discrepancy, and (iii) penalty term. A distributed algorithm is proposed to find the discrepancy as well as the fine tuned neural network at each node. It is experimentally shown that this proposed method outperforms existing algorithms such as FedSGD, DITTO, FedAvg and locally trained neural network with good generalization on various data sets.
{"title":"Multi-Task Federated Edge Learning (MTFeeL) With SignSGD","authors":"Sawan Singh Mahara, M. Shruti, B. Bharath","doi":"10.1109/NCC55593.2022.9806778","DOIUrl":"https://doi.org/10.1109/NCC55593.2022.9806778","url":null,"abstract":"The paper proposes a novel Federated Learning (FL) algorithm involving signed gradient as feedback to reduce communication overhead. The Multi-task nature of the algorithm provides each device a custom neural network after completion. Towards improving the performance, a weighted average loss across devices is proposed which considers the similarity between their data distributions. A Probably Approximately Correct (PAC) bound on the true loss in terms of the proposed empirical loss is derived. The bound is in terms of (i) Rademacher complexity, (ii) discrepancy, and (iii) penalty term. A distributed algorithm is proposed to find the discrepancy as well as the fine tuned neural network at each node. It is experimentally shown that this proposed method outperforms existing algorithms such as FedSGD, DITTO, FedAvg and locally trained neural network with good generalization on various data sets.","PeriodicalId":403870,"journal":{"name":"2022 National Conference on Communications (NCC)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129663608","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：