The last two decades have seen an escalating interest in methods for large-scale unconstrained face recognition. While the promise of computer vision systems to efficiently and accurately verify and identify faces in ...
详细信息
The last two decades have seen an escalating interest in methods for large-scale unconstrained face recognition. While the promise of computer vision systems to efficiently and accurately verify and identify faces in naturally occurring circumstances still remains elusive, recent advances in deep learning are taking us closer to human-level recognition. In this study, the authors propose a new paradigm which employs deep features in a feature extractor and intra-personal factor analysis as a recogniser. The proposed new strategy represents the face changes of a person using identity specific components and the intra-personal variation through reinterpretation of a Bayesian generative factor analysis model. The authors employ the expectation-maximisation algorithm to calculate model parameters which cannot be observed directly. Recognition outcomes achieved through benchmarking on large-scale wild databases, Labeled Faces in the Wild (LFW) and Youtube Face (YTF), clearly prove that the proposed approach provides remarkable face verification performance improvement over state-of-the-art approaches.
The surface of various carbon black and silica grades is characterized via static gas adsorption using different gases. From decomposition of the adsorption isotherm into distinct energetic contributions, the adsorpti...
详细信息
The surface of various carbon black and silica grades is characterized via static gas adsorption using different gases. From decomposition of the adsorption isotherm into distinct energetic contributions, the adsorption energy distribution as well as the surface area are obtained. The decomposition is done by an iterative expectation maximization algorithm specifically designed for this problem. It is found that the adsorption isotherms of the various gases differ significantly in the low-pressure regime, leading to characteristic energy distributions with distinct maxima. As expected, the mean adsorption energy generally increases with the cross section of the gases, and systematic deviations are found reflecting the polar and dispersive interaction characteristics of silica and carbon black, respectively. The surface fractal dimension of two different carbon black grades is estimated using the yardstick method. The obtained values 2.6 and 2.7 agree with previous findings that the carbon black surface morphology is very rough. The adsorption of CO2 on both carbon blacks delivers unexpectedly low values of the monolayer coverage or specific surface area, indicating that mainly high energetic sites of the surface are covered. In consequence, compared with N-2, a relatively high value of the mean adsorption energy is found. For both investigated silicas, the mean adsorption energy scales with the quadrupole moments of CO2 and N-2, which is indicative of a large polar contribution to interaction energy.
In this study, the problem of recovering structured sparse signals with a priori distribution whose structure patterns are unknown is studied from one-bit adaptive (AD) quantised measurements. A generalised approximat...
详细信息
In this study, the problem of recovering structured sparse signals with a priori distribution whose structure patterns are unknown is studied from one-bit adaptive (AD) quantised measurements. A generalised approximate message passing (GAMP) algorithm is utilised, and an expectationmaximisation (EM) method is embedded in the algorithm to iteratively estimate the unknown a priori distribution. In addition, the nearest neighbour sparsity pattern learning (NNSPL) method is adopted to further improve the recovery performance of the structured sparse signals. Numerical results demonstrate the effectiveness of GAMP-EM-AD-NNSPL method with both simulated and real data.
This study deals with the problem of joint delay-Doppler estimation in a practically motivated scenario of passive bistatic radar, where the surveillance channel is polluted by the direct-path signal residual. A new j...
详细信息
This study deals with the problem of joint delay-Doppler estimation in a practically motivated scenario of passive bistatic radar, where the surveillance channel is polluted by the direct-path signal residual. A new joint delay-Doppler maximum-likelihood estimator (MLE) based on Markov chain Monte Carlo (MCMC) is proposed. The MCMC method allows one to compute the MLE in a computationally efficient manner. The proposed estimator is based upon generating random variates using a Markov Chain whose stationary distribution approximates the likelihood function and guarantees convergence to the global maximum. In contrast to the recently proposed modified cross-correlation estimator, and the expectation-maximisation-based MLE, it avoids grid search which may lead to a straddle loss or initialisation-dependent iteration which may lead to convergence problems. Simulation results indicate that the proposed estimator achieves a significant performance improvement over existing methods.
The gradient is an important property in an image. According to the characteristics of image gradient histogram (image gradient magnitude distribution), Gamma distribution model is near to the actual distribution, so ...
详细信息
The gradient is an important property in an image. According to the characteristics of image gradient histogram (image gradient magnitude distribution), Gamma distribution model is near to the actual distribution, so Gamma mixture model is used to fit natural image gradient distribution. First, an image can be divided into edge region and non-edge region on the aspect of the gradient;the authors assume that each region obeys sub-Gamma distribution with different parameters. Then, expectationmaximisation (EM) algorithm is used to estimate the parameters of each part. Finally, the accuracy of the fitting result of the entire gradient distribution is verified by the correlation coefficient and the validity of the estimated gradient magnitude distribution of non-edge region and edge region is verified by the edge-detection experiment with different threshold. This work can select Canny edge-detector high threshold adaptively, which can improve algorithm automatic level.
We propose a novel model to address the task of Visual Dialog which exhibits complex dialog structures. To obtain a reasonable answer based on the current question and the dialog history, the underlying semantic depen...
详细信息
ISBN:
(纸本)9781728132945
We propose a novel model to address the task of Visual Dialog which exhibits complex dialog structures. To obtain a reasonable answer based on the current question and the dialog history, the underlying semantic dependencies between dialog entities are essential. In this paper, we explicitly formalize this task as inference in a graphical model with partially observed nodes and unknown graph structures (relations in dialog). The given dialog entities are viewed as the observed nodes. The answer to a given question is represented by a node with missing value. We first introduce an expectation Maximization algorithm to infer-both the underlying dialog structures and the missing node values (desired answers). Based on this, we proceed to propose a differentiable graph neural network (GNN) solution that approximates this process. Experiment results on the VisDial and VisDial-Q datasets show that our model outperforms comparative methods. It is also observed that our method can infer the underlying dialog structure for better-dialog reasoning.
In this work, we study the robustness of a CNN+RNN based image captioning system being subjected to adversarial noises. We propose to fool an image captioning system to generate some targeted partial captions for an i...
详细信息
ISBN:
(纸本)9781728132945
In this work, we study the robustness of a CNN+RNN based image captioning system being subjected to adversarial noises. We propose to fool an image captioning system to generate some targeted partial captions for an image polluted by adversarial noises, even the targeted captions are totally irrelevant to the image content. A partial caption indicates that the words at some locations in this caption are observed, while words at other locations are not restricted. It is the first work to study exact adversarial attacks of targeted partial captions. Due to the sequential dependencies among words in a caption, we formulate the generation of adversarial noises for targeted partial captions as a structured output learning problem with latent variables. Both the generalized expectation maximization algorithm and structural SVMs with latent variables are then adopted to optimize the problem. The proposed methods generate very successful attacks to three popular CNN+RNN based image captioning models. Furth ermore, the proposed attack methods are used to understand the inner mechanism of image captioning systems, providing the guidance to further improve automatic image captioning systems towards human captioning.
In this paper we develop an identification algorithm to obtain an estimation of the prior distribution in the classical problem of Bayesian inference. We consider the Empirical Bayes approach to obtain the prior distr...
详细信息
ISBN:
(纸本)9781728131863
In this paper we develop an identification algorithm to obtain an estimation of the prior distribution in the classical problem of Bayesian inference. We consider the Empirical Bayes approach to obtain the prior distribution approximation by a finite Gaussian mixture. An expectation-Maximization based algorithm is used to obtain an estimate of the Gaussian mixture parameters. Our approach shows a good approximation of the prior distribution when the number of experiments is increased. We illustrate the estimation performance of our proposal with numerical simulations.
Outlier detection is an important aspect in the field of data mining. In order to solve the problem of outlier detection in high-dimensional datasets, an outlier detection algorithm based on Gaussian mixture model is ...
详细信息
ISBN:
(纸本)9781728137216
Outlier detection is an important aspect in the field of data mining. In order to solve the problem of outlier detection in high-dimensional datasets, an outlier detection algorithm based on Gaussian mixture model is proposed. First of all, for the data set to be tested, the global optimization expectation maximization algorithm is used to fit a Gaussian mixture model, and then the three-time standard deviation principle is introduced on each Gaussian component, the outlier is the data point outside the range of the mean deviation of the mean value of three times the standard deviation. Through the experiments on the simulation dataset and the real data set, the effectiveness of the algorithm on the outlier detection of high-dimensional data sets is verified.
Runtime predictive analysis of quantitative models can support software reliability in various application scenarios. The spread of logging technologies promotes approaches where such models are learned from observed ...
详细信息
ISBN:
(纸本)9781728149837
Runtime predictive analysis of quantitative models can support software reliability in various application scenarios. The spread of logging technologies promotes approaches where such models are learned from observed events. We consider a system visiting transient states of a hidden process until reaching a final state and producing observations with stochastic arrival times and types conditioned by visited states, and we abstract it as a marked Markov modulated Poisson Process (MMMPP) with left-to right structure. We present an expectation-Maximization (EM) algorithm that learns the MMMPP parameters from observation sequences acquired in repeated execution of the transient behavior, and we use the model at runtime to infer the current state of the process from actual observed events and to dynamically evaluate the remaining time to the final state. The approach is illustrated using synthetic datasets generated from a stochastic attack tree of the literature enriched with an observation model associating each state with an expected statistics of observation types and arrival times. Accuracy of prediction is evaluated under different variability of hidden states sojourn durations and of the observations arrival process, and compared against previous literature that mainly exploits either the timing or the types of observed events.
暂无评论