Pub Date : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032149
Chia Wei Lim, M. Wakin
The theory of Compressive Sensing (CS) enables the efficient acquisition of signals which are sparse or compressible in an appropriate domain. In the sub-field of CS known as model-based CS, prior knowledge of the signal sparsity profile is used to improve compression and sparse signal recovery rates. In this paper, we show that by exploiting the periodic support of Periodic Clustered Sparse (PCS) signals, model-based CS improves upon classical CS. We quantify this improvement in terms of simulations performed with a proposed greedy algorithm for PCS signal recovery and provide sampling bounds for the recovery of PCS signals from compressive measurements.
{"title":"Recovery of Periodic Clustered Sparse signals from compressive measurements","authors":"Chia Wei Lim, M. Wakin","doi":"10.1109/GlobalSIP.2014.7032149","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032149","url":null,"abstract":"The theory of Compressive Sensing (CS) enables the efficient acquisition of signals which are sparse or compressible in an appropriate domain. In the sub-field of CS known as model-based CS, prior knowledge of the signal sparsity profile is used to improve compression and sparse signal recovery rates. In this paper, we show that by exploiting the periodic support of Periodic Clustered Sparse (PCS) signals, model-based CS improves upon classical CS. We quantify this improvement in terms of simulations performed with a proposed greedy algorithm for PCS signal recovery and provide sampling bounds for the recovery of PCS signals from compressive measurements.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116294534","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 : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032181
Pablo Peso Parada, D. Sharma, P. Naylor, T. Waterschoot
We present a phoneme confusion analysis that models the impact of reverberation on automatic speech recognition performance by formulating the problem in a Bayesian framework. Our analysis under reverberant conditions shows the relative robustness to reverberation of each phoneme and also indicates that substitutions and deletions correspond to the most common errors in a phoneme recognition task. Finally, a model is proposed to estimate the confusability of each phoneme depending on the reverberation level which is evaluated using two independent data sets.
{"title":"Reverberant speech recognition: A phoneme analysis","authors":"Pablo Peso Parada, D. Sharma, P. Naylor, T. Waterschoot","doi":"10.1109/GlobalSIP.2014.7032181","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032181","url":null,"abstract":"We present a phoneme confusion analysis that models the impact of reverberation on automatic speech recognition performance by formulating the problem in a Bayesian framework. Our analysis under reverberant conditions shows the relative robustness to reverberation of each phoneme and also indicates that substitutions and deletions correspond to the most common errors in a phoneme recognition task. Finally, a model is proposed to estimate the confusability of each phoneme depending on the reverberation level which is evaluated using two independent data sets.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114620928","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 : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032172
Marc Delcroix, Takuya Yoshioka, A. Ogawa, Yotaro Kubo, M. Fujimoto, N. Ito, K. Kinoshita, Miquel Espi, S. Araki, Takaaki Hori, T. Nakatani
Automatic speech recognition is being used successfully in more and more products. However, current recognition systems usually require the use of close-talking microphones. This constraint limits the deployment of speech recognition for new applications. In hands-free situations, noise and reverberation cause a severe degradation of the recognition performance. The problem of noise robustness has attracted a great deal of attention and practical solutions have been proposed and evaluated with common benchmarks. In contrast, reverberation has long been considered an unsolvable problem. Recently, significant progress has been made in the field of reverberant speech recognition and this progress has been evaluated with the REVERB challenge 2014. In this paper, we describe the reverberant speech recognition system we proposed for the REVERB challenge that exhibited high recognition performance even under severe reverberation conditions. We compare our system with other proposed approaches to suggest potential future research directions in the field.
{"title":"Defeating reverberation: Advanced dereverberation and recognition techniques for hands-free speech recognition","authors":"Marc Delcroix, Takuya Yoshioka, A. Ogawa, Yotaro Kubo, M. Fujimoto, N. Ito, K. Kinoshita, Miquel Espi, S. Araki, Takaaki Hori, T. Nakatani","doi":"10.1109/GlobalSIP.2014.7032172","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032172","url":null,"abstract":"Automatic speech recognition is being used successfully in more and more products. However, current recognition systems usually require the use of close-talking microphones. This constraint limits the deployment of speech recognition for new applications. In hands-free situations, noise and reverberation cause a severe degradation of the recognition performance. The problem of noise robustness has attracted a great deal of attention and practical solutions have been proposed and evaluated with common benchmarks. In contrast, reverberation has long been considered an unsolvable problem. Recently, significant progress has been made in the field of reverberant speech recognition and this progress has been evaluated with the REVERB challenge 2014. In this paper, we describe the reverberant speech recognition system we proposed for the REVERB challenge that exhibited high recognition performance even under severe reverberation conditions. We compare our system with other proposed approaches to suggest potential future research directions in the field.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"195 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114748657","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 : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032146
S. Wahls, V. Koivunen, H. Poor, M. Verhaegen
How to learn a function from observations of inputs and noisy outputs is a fundamental problem in machine learning. Often, an approximation of the desired function is found by minimizing a risk functional over some function space. The space of candidate functions should contain good approximations of the true function, but it should also be such that the minimization of the risk functional is computationally feasible. In this paper, finite multidimensional Fourier series are used as candidate functions. Their impressive approximative capabilities are illustrated by showing that Gaussian-kernel estimators can be approximated arbitrarily well over any compact set of bandwidths with a fixed number of Fourier coefficients. However, the solution of the associated risk minimization problem is computationally feasible only if the dimension d of the inputs is small because the number of required Fourier coefficients grows exponentially with d. This problem is addressed by using the tensor train format to model the tensor of Fourier coefficients under a low-rank constraint. An algorithm for least-squares regression is derived and the potential of this approach is illustrated in numerical experiments. The computational complexity of the algorithm grows only linearly both with the number of observations N and the input dimension d, making it feasible also for large-scale problems.
{"title":"Learning multidimensional Fourier series with tensor trains","authors":"S. Wahls, V. Koivunen, H. Poor, M. Verhaegen","doi":"10.1109/GlobalSIP.2014.7032146","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032146","url":null,"abstract":"How to learn a function from observations of inputs and noisy outputs is a fundamental problem in machine learning. Often, an approximation of the desired function is found by minimizing a risk functional over some function space. The space of candidate functions should contain good approximations of the true function, but it should also be such that the minimization of the risk functional is computationally feasible. In this paper, finite multidimensional Fourier series are used as candidate functions. Their impressive approximative capabilities are illustrated by showing that Gaussian-kernel estimators can be approximated arbitrarily well over any compact set of bandwidths with a fixed number of Fourier coefficients. However, the solution of the associated risk minimization problem is computationally feasible only if the dimension d of the inputs is small because the number of required Fourier coefficients grows exponentially with d. This problem is addressed by using the tensor train format to model the tensor of Fourier coefficients under a low-rank constraint. An algorithm for least-squares regression is derived and the potential of this approach is illustrated in numerical experiments. The computational complexity of the algorithm grows only linearly both with the number of observations N and the input dimension d, making it feasible also for large-scale problems.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128506852","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 : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032211
W. Guicquero, P. Vandergheynst, T. Laforest, A. Dupret
Recently developed Compressive Sensing image sensor architectures tend to provide compact on-chip implementations to perform alternative acquisitions. On the other hand, the time of reconstruction generally limits possible applications taking advantage of those specific sensing schemes. This work proposes an entire Compressive Sensing system composed of an encoder (a dedicated imager top-level architecture) and a decoder (a reconstruction algorithm). The proposed system provides a compromise between the sensing scheme efficiency for relaxing on-chip constraints and the reconstruction complexity/quality. This system performs an adaptive block-based sensing, particularly well suited for video acquisition because of being combined with a fast inpainting based reconstruction algorithm. The simulation results show that compared to state of the art reconstructions and without important image degradation, the proposed reconstruction algorithm considerably reduces the computation time.
{"title":"On adaptive pixel random selection for compressive sensing","authors":"W. Guicquero, P. Vandergheynst, T. Laforest, A. Dupret","doi":"10.1109/GlobalSIP.2014.7032211","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032211","url":null,"abstract":"Recently developed Compressive Sensing image sensor architectures tend to provide compact on-chip implementations to perform alternative acquisitions. On the other hand, the time of reconstruction generally limits possible applications taking advantage of those specific sensing schemes. This work proposes an entire Compressive Sensing system composed of an encoder (a dedicated imager top-level architecture) and a decoder (a reconstruction algorithm). The proposed system provides a compromise between the sensing scheme efficiency for relaxing on-chip constraints and the reconstruction complexity/quality. This system performs an adaptive block-based sensing, particularly well suited for video acquisition because of being combined with a fast inpainting based reconstruction algorithm. The simulation results show that compared to state of the art reconstructions and without important image degradation, the proposed reconstruction algorithm considerably reduces the computation time.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"24 3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132829551","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 : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032102
Liwei Yan, B. Bai, Wei Chen
Energy efficiency (EE) maximization for OFDM transceiver has received much attention in next generation wireless communication systems. Recently, there have been a lot of works on EE maximization by balancing the transmission rate and the holistic power consumption. However, how to maximize EE in OFDM systems by joint power allocation and subcarrier selection under holistic power models has not been extensively investigated yet. In contrast to spectrum efficiency (SE) oriented power allocation and subcarrier selection, which is by a simple waterfilling policy, the EE oriented power allocation and subcarrier selection is a mixed integer programming, which is not trivial to solve. This is simply because an extra circuit power has to be paid as the cost of adding an subcarrier. Fortunately, by delving into the mixed integer programming, we found in this paper that the EE oriented power allocation also has form of waterfilling. Based on these results and approximate methods, explicit decision criterions are proposed for subcarrier selection, which provides insight into how the transmit power and the number of subcarriers balance with the channel states and the circuit power in EE oriented systems. Numerical results show that the proposed strategies achieve near-optimal performance of EE with low complexity.
{"title":"Joint power allocation and subcarrier selection for energy efficiency maximization in OFDM systems under a holistic power model","authors":"Liwei Yan, B. Bai, Wei Chen","doi":"10.1109/GlobalSIP.2014.7032102","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032102","url":null,"abstract":"Energy efficiency (EE) maximization for OFDM transceiver has received much attention in next generation wireless communication systems. Recently, there have been a lot of works on EE maximization by balancing the transmission rate and the holistic power consumption. However, how to maximize EE in OFDM systems by joint power allocation and subcarrier selection under holistic power models has not been extensively investigated yet. In contrast to spectrum efficiency (SE) oriented power allocation and subcarrier selection, which is by a simple waterfilling policy, the EE oriented power allocation and subcarrier selection is a mixed integer programming, which is not trivial to solve. This is simply because an extra circuit power has to be paid as the cost of adding an subcarrier. Fortunately, by delving into the mixed integer programming, we found in this paper that the EE oriented power allocation also has form of waterfilling. Based on these results and approximate methods, explicit decision criterions are proposed for subcarrier selection, which provides insight into how the transmit power and the number of subcarriers balance with the channel states and the circuit power in EE oriented systems. Numerical results show that the proposed strategies achieve near-optimal performance of EE with low complexity.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126987008","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 : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032131
Venkat Roy, S. Gishkori, G. Leus
In this paper, a method for the estimation of the spatial rainfall distribution over a specified service area from a limited number of path-averaged rainfall measurements is proposed. The aforementioned problem is formulated as a nonnegativity constrained convex optimization problem with priors that influence both sparsity and clustering properties of the spatial rainfall distribution. The spatial covariance matrix is derived from the climatological variogram model and used to construct a basis for the spatial rainfall vector. A proper selection of the representation basis and the priors that directly relate to the spatial properties of the rainfall guarantee an efficient reconstruction with a low compression rate (fewer measurements).
{"title":"Spatial rainfall mapping from path-averaged rainfall measurements exploiting sparsity","authors":"Venkat Roy, S. Gishkori, G. Leus","doi":"10.1109/GlobalSIP.2014.7032131","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032131","url":null,"abstract":"In this paper, a method for the estimation of the spatial rainfall distribution over a specified service area from a limited number of path-averaged rainfall measurements is proposed. The aforementioned problem is formulated as a nonnegativity constrained convex optimization problem with priors that influence both sparsity and clustering properties of the spatial rainfall distribution. The spatial covariance matrix is derived from the climatological variogram model and used to construct a basis for the spatial rainfall vector. A proper selection of the representation basis and the priors that directly relate to the spatial properties of the rainfall guarantee an efficient reconstruction with a low compression rate (fewer measurements).","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126109557","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 : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032361
Yingying Wang, Yunpeng Cai
Ischemic stroke is one of the leading causes of death and disability worldwide with inflammatory-immune responses in blood and brain damage. To analyze the severity of ischemic stroke, many studies were performed to find biomarkers based on samples from animal brain tissue models. In this work, we used parallel microRNA-mRNA expression profile from rat brain tissues to construct a network based on negative correlation calculation. PageRank algorithm was used to calculate the importance of network nodes. 14 genes were chosen as featured biomarkers. Results showed these genes were significant on biological levels which indicated us that the biomarkers chosen based on animal models may be helpful in stroke diagnosis, etiology and pathogenesis, thus guiding acute treatment and development of new treatments in the future.
{"title":"A network-based analysis of ischemic stroke using parallel microRNA-mRNA expression profiles","authors":"Yingying Wang, Yunpeng Cai","doi":"10.1109/GlobalSIP.2014.7032361","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032361","url":null,"abstract":"Ischemic stroke is one of the leading causes of death and disability worldwide with inflammatory-immune responses in blood and brain damage. To analyze the severity of ischemic stroke, many studies were performed to find biomarkers based on samples from animal brain tissue models. In this work, we used parallel microRNA-mRNA expression profile from rat brain tissues to construct a network based on negative correlation calculation. PageRank algorithm was used to calculate the importance of network nodes. 14 genes were chosen as featured biomarkers. Results showed these genes were significant on biological levels which indicated us that the biomarkers chosen based on animal models may be helpful in stroke diagnosis, etiology and pathogenesis, thus guiding acute treatment and development of new treatments in the future.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123885248","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 : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032082
Yuke Cui, Wei Xu, Hua Zhang, X. You
This paper investigates robust energy efficient beam-forming for multi-cell downlink transmissions. A bounded uncertainty region is considered to model the impairments of channel state information (CSI) available at the base station (BS). We formulate the problem of beamforming optimization by maximizing the worst case energy efficiency (EE) under CSI uncertainties. Due to the non-convex nature of the problem, we resort to instead maximizing a lower bound to the primal objective function. In this way, the problem is casted to a convex semidefinite programming (SDP) under some specific conditions. We accordingly propose an alternating algorithm, which achieves a noticeable performance gain in terms of the worst case EE.
{"title":"Achieving worst case robustness in energy efficient multiuser multicell cooperation systems","authors":"Yuke Cui, Wei Xu, Hua Zhang, X. You","doi":"10.1109/GlobalSIP.2014.7032082","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032082","url":null,"abstract":"This paper investigates robust energy efficient beam-forming for multi-cell downlink transmissions. A bounded uncertainty region is considered to model the impairments of channel state information (CSI) available at the base station (BS). We formulate the problem of beamforming optimization by maximizing the worst case energy efficiency (EE) under CSI uncertainties. Due to the non-convex nature of the problem, we resort to instead maximizing a lower bound to the primal objective function. In this way, the problem is casted to a convex semidefinite programming (SDP) under some specific conditions. We accordingly propose an alternating algorithm, which achieves a noticeable performance gain in terms of the worst case EE.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"62 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122382576","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 : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032221
Z. Shakeri, Haroon Raja, W. Bajwa
This paper addresses the problem of collaborative training of nonlinear classifiers using big, distributed training data. The supervised learning strategy considered in this paper corresponds to data-driven joint learning of a nonlinear transformation that maps the (training) data to a higher-dimensional feature space and a ridge regression based linear classifier in the feature space. The key aspect of this paper, which distinguishes it from related prior work, is that it assumes: (i) the training data are distributed across a number of interconnected sites, and (ii) sizes of the local training data as well as privacy concerns prohibit exchange of individual training samples between sites. The main contribution of this paper is formulation of an algorithm, termed cloud D-KSVD, that reliably, efficiently and collaboratively learns both the nonlinear map and the linear classifier under these constraints. In order to demonstrate the effectiveness of cloud D-KSVD, a number of numerical experiments on the MNIST dataset are also reported in the paper.
{"title":"Dictionary learning based nonlinear classifier training from distributed data","authors":"Z. Shakeri, Haroon Raja, W. Bajwa","doi":"10.1109/GlobalSIP.2014.7032221","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032221","url":null,"abstract":"This paper addresses the problem of collaborative training of nonlinear classifiers using big, distributed training data. The supervised learning strategy considered in this paper corresponds to data-driven joint learning of a nonlinear transformation that maps the (training) data to a higher-dimensional feature space and a ridge regression based linear classifier in the feature space. The key aspect of this paper, which distinguishes it from related prior work, is that it assumes: (i) the training data are distributed across a number of interconnected sites, and (ii) sizes of the local training data as well as privacy concerns prohibit exchange of individual training samples between sites. The main contribution of this paper is formulation of an algorithm, termed cloud D-KSVD, that reliably, efficiently and collaboratively learns both the nonlinear map and the linear classifier under these constraints. In order to demonstrate the effectiveness of cloud D-KSVD, a number of numerical experiments on the MNIST dataset are also reported in the paper.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"10 8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125143618","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：