arXiv subjects

Da Chen

Publications and source records attributed to Da Chen.

At least 55 records · Page 3Linked to original sources

arXiv2021

Active Anomaly Detection with Switching Cost

The problem of detecting a single anomalous process among multiple independent processes is considered. Under a constraint on the number of processes that can be probed simultaneously, the decision maker should decide which processes to probe at each time and when to terminate the probing. Compared with previous work considering only the observation costs, the switching costs of switchings across processes also need to be taken into account in many practical scenarios. The objective is an active inference strategy that minimizes the Bayesian risk taking into account of the sample complexity, switching cost, as well as detection errors. Based on the framework of sequential design of experiments, we propose a low-complexity, low-switching deterministic policy for two scenarios where the total switching cost is negligible and the total switching cost is comparable to the total observation cost. We show that the proposed algorithm is asymptotically optimal in the former scenario and is order optimal in the latter scenario. Simulation results demonstrate strong performance in the finite regime for both scenarios.

eess.SP↗

arXiv2020

Semantic Regularization: Improve Few-shot Image Classification by Reducing Meta Shift

Few-shot image classification requires the classifier to robustly cope with unseen classes even if there are only a few samples for each class. Recent advances benefit from the meta-learning process where episodic tasks are formed to train a model that can adapt to class change. However, these task sare independent to each other and existing works mainly rely on limited samples of individual support set in a single meta task. This strategy leads to severe meta shift issues across multiple tasks, meaning the learned prototypes or class descriptors are not stable as each task only involves their own support set. To avoid this problem, we propose a concise Semantic RegularizationNetwork to learn a common semantic space under the framework of meta-learning. In this space, all class descriptors can be regularized by the learned semantic basis, which can effectively solve the meta shift problem. The key is to train a class encoder and decoder structure that can encode the sample embedding features into the semantic domain with trained semantic basis, and generate a more stable and general class descriptor from the decoder. We evaluate our work by extensive comparisons with previous methods on three benchmark datasets (MiniImageNet, TieredImageNet, and CUB). The results show that the semantic regularization module improves performance by 4%-7% over the baseline method, and achieves competitive results over the current state-of-the-art models.

cs.CV↗

arXiv2020

Data-Free Adversarial Distillation

Knowledge Distillation (KD) has made remarkable progress in the last few years and become a popular paradigm for model compression and knowledge transfer. However, almost all existing KD algorithms are data-driven, i.e., relying on a large amount of original training data or alternative data, which is usually unavailable in real-world scenarios. In this paper, we devote ourselves to this challenging problem and propose a novel adversarial distillation mechanism to craft a compact student model without any real-world data. We introduce a model discrepancy to quantificationally measure the difference between student and teacher models and construct an optimizable upper bound. In our work, the student and the teacher jointly act the role of the discriminator to reduce this discrepancy, when a generator adversarially produces some "hard samples" to enlarge it. Extensive experiments demonstrate that the proposed data-free method yields comparable performance to existing data-driven methods. More strikingly, our approach can be directly extended to semantic segmentation, which is more complicated than classification, and our approach achieves state-of-the-art results. Code and pretrained models are available at https://github.com/VainF/Data-Free-Adversarial-Distillation.

cs.LG↗

arXiv2019

Spectral Efficiency Analysis of Cell-Free Massive MIMO Systems with Zero-Forcing Detector

In this paper, we firstly derive two approximations of the achievable uplink rate with the perfect/imperfect channel state information (CSI) in cell-free massive multi-input multi-output (MIMO) systems, and all these approximations are not only in the simple, but also converge into the classical bounds achieved in conventional massive MIMO systems where the base-station (BS) antennas are co-located. It is worth noting that the obtained two approximations with perfect CSI could be regarded as the special cases of the obtained two approximations with imperfect CSI when the pilot sequence power becomes infinite, respectively. Moreover, the theory analysis shows that all obtained approximations with perfect/imperfect CSI have an asymptotic lower bound $\fracα{2}\log_2 L$ thanks to the extra {\emph {distance diversity}} offered by massively distributed antennas, where $L$ is the number of BS antennas and the path-loss factor $α>2$, except for the free space environment. Obviously, these results indicate that the cell-free massive MIMO system has huge potential of spectral efficiency than the conventional massive MIMO system with the asymptotically tight bound $\log_2 L$.

cs.IT↗

arXiv2019

A Covariance-Based Hybrid Channel Feedback in FDD Massive MIMO Systems

In this paper, a novel covariance-based channel feedback mechanism is investigated for frequency division duplexing (FDD) massive multi-input multi-output (MIMO) systems. The concept capitalizes on the notion of user statistical separability which was hinted in several prior works in the massive antenna regime but not fully exploited so far. We here propose a hybrid statistical-instantaneous feedback mechanism where the users are separated into two classes of feedback design based on their channel covariance. Under the hybrid framework, each user either operates on a statistical feedback mode or quantized instantaneous channel feedback mode depending on their so-called statistical isolability. The key challenge lies in the design of a covariance-aware classification algorithm which can handle the complex mutual interactions between all users. The classification is derived from rate bound principles. A suitable precoding method is also devised under the mixed statistical and instantaneous feedback model. Simulations are performed to validate our analytical results and illustrate the sum rate advantages of the proposed feedback scheme under a global feedback overhead constraint.

cs.IT↗

arXiv2019

From Active Contours to Minimal Geodesic Paths: New Solutions to Active Contours Problems by Eikonal Equations

In this chapter, we give an overview of part of our previous work based on the minimal path framework and the Eikonal partial differential equation (PDE). We show that by designing adequate Riemannian and Randers geodesic metrics the minimal paths can be utilized to search for solutions to almost all of the active contour problems and to the Euler-Mumford elastica problem, which allows to blend the advantages from minimal geodesic paths and those original approaches, i.e. the active contours and elastica curves. The proposed minimal path-based models can be applied to deal with a broad variety of image analysis tasks such as boundary detection, image segmentation and tubular structure extraction. The numerical implementations for the computation of minimal paths are known to be quite efficient thanks to the Eikonal solvers such as the Finsler variant of the fast marching method.

cs.CV↗

arXiv2019

Millimeter-Wave Massive MIMO Systems Relying on Generalized Sub-Array-Connected Hybrid Precoding

In this paper, we consider a generalized sub-array-connected (GSAC) architecture for arbitrary radio frequency (RF) chain and antenna configurations, where the number of RF chains connected to a sub-array and the number of antennas in each sub-array can be arbitrary. Our design objective is to improve the energy-efficiency of the hybrid precoder of millimeter-wave massive multiple input multiple output (MIMO) systems. We firstly propose a successive interference cancellation based hybrid precoding algorithm to maximize the achievable rate for any given RF chain and antenna configuration. This algorithm firstly decomposes the total achievable rate optimization problem into multiple sub-rate optimization problems, then it successively maximizes these sub-rates. Since the number of RF chains is limited, we can afford using an exhaustive search scheme to evaluate all configurations and identify the one having the best energy efficiency. Moreover, to rely on an attractive limited feedback, we also propose a beamsteering codebook for our hybrid precoding aided GSAC architecture. Our simulation results demonstrate that the proposed scheme achieves a similar rate as the corresponding optimal unconstrained precoder. Furthermore, we show that the energy-efficiency of the proposed scheme is better than that of the existing schemes in the fully-connected and sub-array-connected architectures.

cs.IT↗

arXiv2019

KPI/KQI-Driven Coordinated Multi-Point in 5G: Measurements, Field Trials, and Technical Solutions

The fifth generation (5G) systems are expected to be able to support massive number of wireless devices and intense demands for high data rates while maintaining low latency. Coordinated multipoint (CoMP) is advocated by recent advances and is envisioned to continue its adoption in 5G to meet these requirements by alleviating inter-cell interference and improving spectral efficiency. The higher requirements in 5G have raised the stakes on developing a new CoMP architecture. To understand the merits and limitations of CoMP in 5G, this article systematically investigates evaluation criteria including key performance indicators (KPIs) and key quality indicators (KQIs) in 5G, conducts empirical measurements and field tests, and then proposes a KPI/KQI-driven CoMP architecture that fulfills KPI requirements and provides KQI guarantee for each user.

cs.NI↗

arXiv2019

Deep Features Analysis with Attention Networks

Deep neural network models have recently draw lots of attention, as it consistently produce impressive results in many computer vision tasks such as image classification, object detection, etc. However, interpreting such model and show the reason why it performs quite well becomes a challenging question. In this paper, we propose a novel method to interpret the neural network models with attention mechanism. Inspired by the heatmap visualization, we analyze the relation between classification accuracy with the attention based heatmap. An improved attention based method is also included and illustrate that a better classifier can be interpreted by the attention based heatmap.

cs.LG↗

arXiv2018

Spatial Lobes Division Based Low Complexity Hybrid Precoding and Diversity Combining for mmWave IoT Systems

This paper focuses on the design of low complexity hybrid analog/digital precoding and diversity combining in millimeter wave (mmWave) Internet of things (IoT) systems. Firstly, by exploiting the sparseness property of the mmWave in the angular domain, we propose a spatial lobes division (SLD) to group the total paths of the mmWave channel into several spatial lobes, where the paths in each spatial lobe form a low-rank sub-channel. Secondly, based on the SLD operation, we propose a low complexity hybrid precoding scheme, named HYP-SLD. Specifically, for each low-rank sub-channel, we formulate the hybrid precoding design as a sparse reconstruction problem and separately maximizes the spectral efficiency. Finally, we further propose a maximum ratio combining based diversity combining scheme, named HYP-SLD-MRC, to improve the bit error rate (BER) performance of mmWave IoT systems. Simulation results demonstrate that, the proposed HYP-SLD scheme significantly reduces the complexity of the classic orthogonal matching pursuit (OMP) scheme. Moreover, the proposed HYP-SLD-MRC scheme achieves great improvement in BER performance compared with the fully digital precoding scheme.

cs.IT↗

arXiv2018

Minimal Paths for Tubular Structure Segmentation with Coherence Penalty and Adaptive Anisotropy

The minimal path method has proven to be particularly useful and efficient in tubular structure segmentation applications. In this paper, we propose a new minimal path model associated with a dynamic Riemannian metric embedded with an appearance feature coherence penalty and an adaptive anisotropy enhancement term. The features that characterize the appearance and anisotropy properties of a tubular structure are extracted through the associated orientation score. The proposed dynamic Riemannian metric is updated in the course of the geodesic distance computation carried out by the efficient single-pass fast marching method. Compared to state-of-the-art minimal path models, the proposed minimal path model is able to extract the desired tubular structures from a complicated vessel tree structure. In addition, we propose an efficient prior path-based method to search for vessel radius value at each centerline position of the target. Finally, we perform the numerical experiments on both synthetic and real images. The quantitive validation is carried out on retinal vessel images. The results indicate that the proposed model indeed achieves a promising performance.

cs.CV↗

arXiv2018

Spectral Efficiency Analysis of Multi-Cell Massive MIMO Systems with Ricean Fading

This paper investigates the spectral efficiency of multi-cell massive multiple-input multiple-output systems with Ricean fading that utilize the linear maximal-ratio combining detector. We firstly present closed-form expressions for the effective signal-to-interference-plus-noise ratio (SINR) with the least squares and minimum mean squared error (MMSE) estimation methods, respectively, which apply for any number of base-station antennas $M$ and any Ricean $K$-factor. Also, the obtained results can be particularized in Rayleigh fading conditions when the Ricean $K$-factor is equal to zero. In the following, novel exact asymptotic expressions of the effective SINR are derived in the high $M$ and high Ricean $K$-factor regimes. The corresponding analysis shows that pilot contamination is removed by the MMSE estimator when we consider both infinite $M$ and infinite Ricean $K$-factor, while the pilot contamination phenomenon persists for the rest of cases. All the theoretical results are verified via Monte-Carlo simulations.

cs.IT↗

arXiv2018

Smoothed SVD-based Beamforming for FBMC/OQAM Systems Based on Frequency Spreading

The combination of singular value decomposition (SVD)-based beamforming and filter bank multicarrier with offset quadrature amplitude modulation (FBMC/OQAM) has not been successful to date. The difficulty of this combination is that, the beamformers may experience significant changes between adjacent subchannels, therefore destroy the orthogonality among FBMC/OQAM real-valued symbols, even under channels with moderate frequency selectivity. In this paper, we address this problem from two aspects: i) an SVD-FS-FBMC architecture is adopted to support beamforming with finer granularity in frequency domain, based on the frequency spreading FBMC (FS-FBMC) structure, i.e., beamforming on FS-FBMC tones rather than on subchannels; ii) criterion and methods are proposed to smooth the beamformers from tone to tone. The proposed finer beamforming and smoothing greatly improve the smoothness of beamformers, therefore effectively suppress the leaked ICI/ISI. Simulations are conducted under the scenario of IEEE 802.11n wireless LAN. Results show that the proposed SVD-FS-FBMC system shares close BER performance with its orthogonal frequency division multiplexing (OFDM) counterpart under the frequency selective channels.

eess.SP↗

arXiv2018

Global Minimum for a Finsler Elastica Minimal Path Approach

In this paper, we propose a novel curvature-penalized minimal path model via an orientation-lifted Finsler metric and the Euler elastica curve. The original minimal path model computes the globally minimal geodesic by solving an Eikonal partial differential equation (PDE). Essentially, this first-order model is unable to penalize curvature which is related to the path rigidity property in the classical active contour models. To solve this problem, we present an Eikonal PDE-based Finsler elastica minimal path approach to address the curvature-penalized geodesic energy minimization problem. We were successful at adding the curvature penalization to the classical geodesic energy. The basic idea of this work is to interpret the Euler elastica bending energy via a novel Finsler elastica metric that embeds a curvature penalty. This metric is non-Riemannian, anisotropic and asymmetric, and is defined over an orientation-lifted space by adding to the image domain the orientation as an extra space dimension. Based on this orientation lifting, the proposed minimal path model can benefit from both the curvature and orientation of the paths. Thanks to the fast marching method, the global minimum of the curvature-penalized geodesic energy can be computed efficiently. We introduce two anisotropic image data-driven speed functions that are computed by steerable filters. Based on these orientation-dependent speed functions, we can apply the proposed Finsler elastica minimal path model to the applications of closed contour detection, perceptual grouping and tubular structure extraction. Numerical experiments on both synthetic and real images show that these applications of the proposed model indeed obtain promising results.

cs.CG↗

arXiv2018

Hybrid Precoding Based on Non-Uniform Quantization Codebook to Reduce Feedback Overhead in Millimeter Wave MIMO Systems

In this paper, we focus on the design of the hybrid analog/digital precoding in millimeter wave multiple-input multiple-output (MIMO) systems. To reduce the feedback overhead, we propose two non-uniform quantization (NUQ) codebook based hybrid precoding schemes for two main hybrid precoding implementations, i.e., the full-connected structure and the sub-connected structure. Specifically, we firstly group the angles of the arrive/departure (AOAs/AODs) of the scattering paths into several spatial lobes by exploiting the sparseness property of the millimeter wave in the angular domain, which divides the total angular domain into effective spatial lobes' coverage angles and ineffective coverage angles. Then, we map the quantization bits non-uniformly to different coverage angles and construct NUQ codebooks, where high numbers of quantization bits are employed for the effective coverage angles to quantize AoAs/AoDs and zero quantization bit is employed for ineffective coverage angles. Finally, two low-complexity hybrid analog/digital precoding schemes are proposed based on NUQ codebooks. Simulation results demonstrate that, the proposed two NUQ codebook based hybrid precoding schemes achieve near-optimal spectral efficiencies and show the superiority in reducing the feedback overhead compared with the uniform quantization (UQ) codebook based works, e.g., at least 12.5% feedback overhead could be reduced for a system with 144/36 transmitting/receiving antennas.

cs.IT↗

arXiv2018

FDM-Structured Preamble Optimization for Channel Estimation in MIMO-OQAM/FBMC Systems

In this paper, we consider the problem of preamble design in multiple-input multiple-output (MIMO) systems employing offset quadrature amplitude modulation based filter bank multicarrier (OQAM/FBMC) and propose a preamble optimization method for the frequency division multiplexing (FDM)-structured preamble. Specifically, we formulate an optimization problem to determine the frequency division multiplexed preambles, where the objective is to minimize the mean square error (MSE) of the channel estimation, subject to the constraint on the transmit energy. For two transmit antennas, we find the relationship between preambles and the intrinsic interference from neighboring symbols to achieve the minimum channel estimation MSE, and derive the optimal closed-form solution. For more than two transmit antennas, the constrained preamble optimization problem is nonconvex quadratic. Therefore, we convert the original optimization problem into a quadratically constrained quadratic program (QCQP) and obtain the suboptimal solution by relaxing the nonconvex constraint. Simulation results demonstrate that, in terms of MSE and bit error rate (BER) performances, the proposed method outperforms the conventional FDM preamble design method at all signal-to-noise ratio (SNR) regimes and outperforms the interference approximation method-complex (IAM-C) preamble design method at low to medium SNR regimes with lower preamble overhead.

cs.IT↗

arXiv2017

Fast Asymmetric Fronts Propagation for Image Segmentation

In this paper, we introduce a generalized asymmetric fronts propagation model based on the geodesic distance maps and the Eikonal partial differential equations. One of the key ingredients for the computation of the geodesic distance map is the geodesic metric, which can govern the action of the geodesic distance level set propagation. We consider a Finsler metric with the Randers form, through which the asymmetry and anisotropy enhancements can be taken into account to prevent the fronts leaking problem during the fronts propagation. These enhancements can be derived from the image edge-dependent vector field such as the gradient vector flow. The numerical implementations are carried out by the Finsler variant of the fast marching method, leading to very efficient interactive segmentation schemes. We apply the proposed Finsler fronts propagation model to image segmentation applications. Specifically, the foreground and background segmentation is implemented by the Voronoi index map. In addition, for the application of tubularity segmentation, we exploit the level set lines of the geodesic distance map associated to the proposed Finsler metric providing that a thresholding value is given.

cs.CG↗

arXiv2017

Downlink Precoding with Mixed Statistical and Imperfect Instantaneous CSI for Massive MIMO Systems

In this paper, the feasibility of a new downlink transmission mode in massive multi-input multi-output (MIMO) systems is investigated with two types of users, i.e., the users with only statistical channel state information (CSI) and the users with imperfect instantaneous CSI. The problem of downlink precoding design with mixed utilization of statistical and imperfect instantaneous CSI is addressed. We first theoretically analyze the impact of the mutual interference between the two types of users on their achievable rate. Then, considering the mutual interference suppression, we propose an extended zero-forcing (eZF) and an extended maximum ratio transmission (eMRT) precoding methods to minimize the total transmit power of base station and to maximize the received signal power of users, respectively. Thanks to the exploitation of statistical CSI, pilot-based channel estimation is avoided enabling more active users, higher system sum rate and shorter transmission delay. Finally, simulations are performed to validate the accuracy of the theoretical analysis and the advantages of the proposed precoding methods.

cs.IT↗