无监督行人重识别的判别性特征研究

目前,在计算机视觉方面,大多的监督学习方法用于解决其重要分支:行人重识别问题已经取得了不错的成果,但是此类方法需要对训练数据进行手工标注,特别是对于大容量的数据集,手工标注的成本很高,而且完全满足成对标记的数据难以获得,所以无监督学习成为必选项.此外,全局特征注重行人特征空间整体性的判别性,而局部特征有助于凸显不同部位特征的判别性.所以,基于全局与局部特征的无监督学习框架,使用全局损失函数与局部相斥损失函数共同进行判别性特征学习,并联合优化ResNet-50卷积神经网络(CNN)和各个样本之间的关系,最终实现行人重识别.大量实验数据验证了提出的方法在解决行人重识别任务时具有优越性.     

Channel estimation techniques for linear precoded systems: Supervised, unsupervised, and hybrid approaches

Linear precoding is an attractive technique to combat interference in multiple-input multiple-output systems because it reduces cost and power consumption at the receiver. Frequency division duplex systems with linear precoding acquire the channel state information at the receiver side by using supervised algorithms. Such methods make use of pilot symbols periodically provided by the transmitter. Next, this channel state information is sent to the transmitter side through a low-cost feedback channel. Thus, the available channel information allows the transmitter to adapt signals to the channel conditions. Given that pilot symbols do not convey user data, they penalize throughput, spectral efficiency, and transmission energy consumption of the system. In this work, we propose to mitigate the aforementioned limitations by combining both supervised and unsupervised algorithms to acquire the channel state information needed by the transmitter. The key idea consists in introducing a simple criterion to determine whether the channel has suffered a significant variation which requires the transmission of pilot symbols. Otherwise, when small fluctuations happen, an unsupervised method is used to track these channel variations instead. This criterion will be evaluated by considering two types of strategies for the design of the linear precoders: Zero-Forcing and Wiener criteria.     

An analog floating-gate node for Supervised learning

We present an improved analog floating-gate pFET synapse that implements a supervised learning algorithm similar to the least mean square (LMS) learning rule. Weight decay plays a key role in several learning rules; this floating-gate synapse exhibits this behavior. We examine implications of the weight decay appearing in the correlation learning rule realized in the floating-gate synapse and provide experimental data characterizing the synapse and its performance in one-input and two-input LMS networks. Analog floating-gate synapses will enable larger-scale, on-chip learning networks than previously possible.     

Applying radial basis functions

Discusses the application of neural networks to general and radial basis functions and in particular to adaptive equalization and interference rejection problems. Neural-network-based algorithms strike a good balance between performance and complexity in adaptive equalization, and show promise in spread spectrum systems     

Optimal unsupervised learning multicategory dependent hypotheses pattern recognition

A Bayesian decision theory approach is applied to the solution of the problem of unsupervised parametric pattern recognition. The parametric model for this investigation includes the cases where both constant and time-varying unknown parameters are present, and, most significantly, the unknown hypotheses do not constitute a statistically independent sequence. They are restricted only to be from a source with finite-order Markov dependence. The resulting optimal learning system is found and shown to grow initially in size and memory until theNth observation (whereNis the highest Markov order), and subsequently to remain of fixed size and memory. It can, therefore, operate indefinitely and continue to improve its ability to recognize patterns utilizing only a fixed size memory. In summary, the main contributions of this paper are the following: begin{enumerate} item the extension of previous investigations of the unsupervised parametric pattern recognition problem to include cases where both constant and time-varying unknown parameter vectors are simultaneously present; item that the a priori probabilities of the hypotheses, the time-varying parameters, and their transition laws may, if constant, be expressed as functions of the constant unknown parameter and, thus, also be learned; and item the removal of the assumption of statistical independence between hypotheses for the sequence of observations. end{enumerate}     

Real-time data-reusing adaptive learning of a radial basis function network for tracking evoked potentials

Tracking variations in both the latency and amplitude of evoked potential (EP) is important in quantifying properties of the nervous system. Adaptive filtering is a powerful tool for tracking such variations. In this paper, a data-reusing non-linear adaptive filtering method, based on a radial basis function network (RBFN), is implemented to estimate EP. The RBFN consists of an input layer of source nodes, a single hidden layer of non-linear processing units and an output layer of linear weights. It has built-in nonlinear activation functions that allow learning of function mappings. Moreover, it produces satisfactory estimates of signals against a background noise without a priori knowledge of the signal, provided that the signal and noise are independent. In clinical situations where EP responses change rapidly, the convergence rate of the algorithm becomes a critical factor. A carefully designed data-reusing RBFN can accelerate the convergence rate markedly and, thus, enhance its performance. Both theoretical analysis and simulation results support the improved performance of our new algorithm.     

Supervised within-class-similar discriminative dictionary learning for face recognition

The current study puts forward a supervised within-class-similar discriminative dictionary learning (SCDDL) algorithm for face recognition. Some popular discriminative dictionary learning schemes for recognition tasks always incorporate the linear classification error term into the objective function or make some discriminative restrictions on representation coefficients. In the presented SCDDL algorithm, we propose to directly restrict the representation coefficients to be similar within the same class and simultaneously include the linear classification error term in the supervised dictionary learning scheme to derive a more discriminative dictionary for face recognition. The experimental results on three large well-known face databases suggest that our approach can enhance the fisher ratio of representation coefficients when compared with several dictionary learning algorithms that incorporate linear classifiers. In addition, the learned discriminative dictionary, the large fisher ratio of representation coefficients and the simultaneously learned classifier can improve the recognition rate compared with some state-of-the-art dictionary learning algorithms.     

基于非监督学习的恶意欠费用电客户识别

恶意欠费用户识别对供电企业有重要意义,有利于供电企业提前介入,避免电费回收风险。由于用户电费数据缺少标签以区分用户是否恶意欠费,因此需要采取非监督学习方式建立识别模型。文中借鉴信息安全风险评估的风险值概念,综合欠费的恶意程度和欠费造成的经济损失程度来评估用户的欠费风险值,建立基于经验阀值的动态更新的恶意欠费用户识别模型。案例分析表明,该模型具备良好的识别能力,符合人们对恶意欠费用户的经验理解。     

CLUE: cluster-based retrieval of images by unsupervised learning.

In a typical content-based image retrieval (CBIR) system, target images (images in the database) are sorted by feature similarities with respect to the query. Similarities among target images are usually ignored. This paper introduces a new technique, cluster-based retrieval of images by unsupervised learning (CLUE), for improving user interaction with image retrieval systems by fully exploiting the similarity information. CLUE retrieves image clusters by applying a graph-theoretic clustering algorithm to a collection of images in the vicinity of the query. Clustering in CLUE is dynamic. In particular, clusters formed depend on which images are retrieved in response to the query. CLUE can be combined with any real-valued symmetric similarity measure (metric or nonmetric). Thus, it may be embedded in many current CBIR systems, including relevance feedback systems. The performance of an experimental image retrieval system using CLUE is evaluated on a database of around 60,000 images from COREL. Empirical results demonstrate improved performance compared with a CBIR system using the same image similarity measure. In addition, results on images returned by Google's Image Search reveal the potential of applying CLUE to real-world image data and integrating CLUE as a part of the interface for keyword-based image retrieval systems.     

A quasi-Bayes unsupervised learning procedure for priors (Corresp.)

Unsupervised Bayes sequential learning procedures for classification and estimation are often useless in practice because of the amount of computation required. In this paper, a version of a two-class decision problem is considered, and a quasi-Bayes procedure is motivated and defined. The proposed procedure mimics closely the formal Bayes solution while involving only a minimal amount of computation. Convergence properties are established and some numerical illustrations provided. The approach compares favorably with other non-Bayesian learning procedures that have been proposed and can be extended to more general situations.     

Online unsupervised learning of hidden Markov models for adaptivespeech recognition

A novel framework of an online unsupervised learning algorithm is presented to flexibly adapt the existing speaker-independent hidden Markov models (HMMs) to nonstationary environments induced by varying speakers, transmission channels, ambient noises, etc. The quasi-Bayes (QB) estimate is applied to incrementally obtain word sequence and adaptation parameters for adjusting HMMs when a block of unlabelled data is enrolled. The underlying statistics of a nonstationary environment can be successively traced according to the newest enrolment data. To improve the QB estimate, the adaptive initial hyperparameters are employed in the beginning session of online learning. These hyperparameters are estimated from a cluster of training speakers closest to the test environment. Additionally, a selection process is developed to select reliable parameters from a list of candidates for unsupervised learning. A set of reliability assessment criteria is explored for selection. In a series of speaker adaptation experiments, the effectiveness of the proposed method is confirmed and it is found that using the adaptive initial hyperparameters in online learning and the multiple assessments in parameter selection can improve the recognition performance     

Gaussian radial basis functions and inner product spaces

An approximation result is given concerning Gaussian radial basis functions in a general inner product space. Applications are described concerning the classification of the elements of disjoint sets of signals, and also the approximation of continuous real functions defined on all of ⁿ using radial basis function (RBF) networks. More specifically, it is shown that an important large class of classification problems involving signals can be solved using a structure consisting of only a generalized RBF network followed by a quantizer. It is also shown that Gaussian radial basis functions defined on ⁿ can uniformly approximate arbitrarily well over all of ⁿ any continuous real functionalf on ⁿ that meets the condition that |f(x)|0 as x.     

Multilayer perceptron trained using radial basis functions

An algebraic equation for the training of a multilayer perceptron using radial basis functions is derived. Examples using this technique for the training of a network for the exclusive-OR and related problems are presented. Suggestions on the choice of the number of centres are given.<>     

DCT basis function learning control

In this paper, a basis function learning control is developed in away parallel to time domain learning control. Basis function approach aims to reduce the dimension and computation of learning gain matrices while maintaining minimal loss of tracking precision. Here, we transplant two learning gain matrices, the transpose and the partial isometry, from time domain learning control into basis function learning control. These two learning gain matrices have no ill-conditioned problems in matrix computation and ensure a monotonic decay of tracking error. The basis vector of discrete cosine transform (DCT) is chosen as basis function for its high energy compression ratio and energy preservation feature. Experiments on two joints of a SCARA type robot verify the effectiveness of the proposed approaches. A few DCT coefficients may meet learning control specification and tracking precision can be improved by increasing the number of the DCT coefficients.     

Flood forecasting using radial basis function neural networks

A radial basis function (RBF) neural network (NN) is proposed to develop a rainfall-runoff model for three-hour-ahead flood forecasting. For faster training speed, the RBF NN employs a hybrid two-stage learning scheme. During the first stage, unsupervised learning, fuzzy min-max clustering is introduced to determine the characteristics of the nonlinear RBFs. In the second stage, supervised learning, multivariate linear regression is used to determine the weights between the hidden and output layers. The rainfall-runoff relation can be considered as a linear combination of some nonlinear RBFs. Rainfall and runoff events of the Lanyoung River collected during typhoons are used to train, validate,and test the network. The results show that the RBF NN can be considered a suitable technique for predicting flood flow     

Channel equalization using adaptive complex radial basis functionnetworks

It is generally recognized that digital channel equalization can be interpreted as a problem of nonlinear classification. Networks capable of approximating nonlinear mappings can be quite useful in such applications. The radial basis function network (RBFN) is one such network. We consider an extension of the RBFN for complex-valued signals (the complex RBFN or CRBFN). We also propose a stochastic-gradient (SG) training algorithm that adapts all free parameters of the network. We then consider the problem of equalization of complex nonlinear channels using the CRBFN as part of an equalizer. Results of simulations we have carried out show that the CRBFN with the SG algorithm can be quite effective in channel equalization     

Burst-by-burst adaptive turbo-coded radial basis function-assisteddecision feedback equalization

The performance of the proposed radial basis function (RBF) assisted turbo-coded adaptive modulation scheme is characterized in a wideband channel scenario. We commence by introducing the novel concept of the Jacobian RBF equalizer, which is a reduced-complexity version of the conventional RBF equalizer. Specifically, the Jacobian logarithmic RBF equalizer generates its output in the logarithmic domain and hence it can be used to provide soft outputs for the turbo-channel decoder. We propose using the average magnitude of the log-likelihood ratio (LLR) of the bits in the received transmission burst before channel decoding as the channel quality measure for controlling the mode-switching regime of our adaptive scheme     

Generative-discriminative basis learning for medical imaging

This paper presents a novel dimensionality reduction method for classification in medical imaging. The goal is to transform very high-dimensional input (typically, millions of voxels) to a low-dimensional representation (small number of constructed features) that preserves discriminative signal and is clinically interpretable. We formulate the task as a constrained optimization problem that combines generative and discriminative objectives and show how to extend it to the semi-supervised learning (SSL) setting. We propose a novel large-scale algorithm to solve the resulting optimization problem. In the fully supervised case, we demonstrate accuracy rates that are better than or comparable to state-of-the-art algorithms on several datasets while producing a representation of the group difference that is consistent with prior clinical reports. Effectiveness of the proposed algorithm for SSL is evaluated with both benchmark and medical imaging datasets. In the benchmark datasets, the results are better than or comparable to the state-of-the-art methods for SSL. For evaluation of the SSL setting in medical datasets, we use images of subjects with mild cognitive impairment (MCI), which is believed to be a precursor to Alzheimer's disease (AD), as unlabeled data. AD subjects and normal control (NC) subjects are used as labeled data, and we try to predict conversion from MCI to AD on follow-up. The semi-supervised extension of this method not only improves the generalization accuracy for the labeled data (AD/NC) slightly but is also able to predict subjects which are likely to converge to AD.     

Sparse Bayesian learning for basis selection

Sparse Bayesian learning (SBL) and specifically relevance vector machines have received much attention in the machine learning literature as a means of achieving parsimonious representations in the context of regression and classification. The methodology relies on a parameterized prior that encourages models with few nonzero weights. In this paper, we adapt SBL to the signal processing problem of basis selection from overcomplete dictionaries, proving several results about the SBL cost function that elucidate its general behavior and provide solid theoretical justification for this application. Specifically, we have shown that SBL retains a desirable property of the /spl lscr//sub 0/-norm diversity measure (i.e., the global minimum is achieved at the maximally sparse solution) while often possessing a more limited constellation of local minima. We have also demonstrated that the local minima that do exist are achieved at sparse solutions. Later, we provide a novel interpretation of SBL that gives us valuable insight into why it is successful in producing sparse representations. Finally, we include simulation studies comparing sparse Bayesian learning with basis pursuit and the more recent FOCal Underdetermined System Solver (FOCUSS) class of basis selection algorithms. These results indicate that our theoretical insights translate directly into improved performance.