Out-of-distribution (OOD) detection has recently gained substantial attention due to the importance of identifying out-of-domain samples in reliability and safety. Although OOD detection methods have advanced by a great deal, they are still susceptible to adversarial examples, which is a violation of their purpose. To mitigate this issue, several defenses have recently been proposed. Nevertheless, these efforts remained ineffective, as their evaluations are based on either small perturbation sizes, or weak attacks. In this work, we re-examine these defenses against an end-to-end PGD attack on in/out data with larger perturbation sizes, e.g. up to commonly used $\epsilon=8/255$ for the CIFAR-10 dataset. Surprisingly, almost all of these defenses perform worse than a random detection under the adversarial setting. Next, we aim to provide a robust OOD detection method. In an ideal defense, the training should expose the model to almost all possible adversarial perturbations, which can be achieved through adversarial training. That is, such training perturbations should based on both in- and out-of-distribution samples. Therefore, unlike OOD detection in the standard setting, access to OOD, as well as in-distribution, samples sounds necessary in the adversarial training setup. These tips lead us to adopt generative OOD detection methods, such as OpenGAN, as a baseline. We subsequently propose the Adversarially Trained Discriminator (ATD), which utilizes a pre-trained robust model to extract robust features, and a generator model to create OOD samples. We noted that, for the sake of training stability, in the adversarial training of the discriminator, one should attack real in-distribution as well as real outliers, but not generated outliers. Using ATD with CIFAR-10 and CIFAR-100 as the in-distribution data, we could significantly outperform all previous methods in the robust AUROC while maintaining high standard AUROC and classification accuracy. The code repository is available at https://github.com/rohban-lab/ATD.

It is of fundamental importance to find algorithms obtaining optimal performance for learning of statistical models in distributed and communication limited systems. Aiming at characterizing the optimal strategies, we consider learning of Gaussian Processes (GP) in distributed systems as a pivotal example. We first address a very basic problem: how many bits are required to estimate the inner-products of some Gaussian vectors across distributed machines? Using information theoretic bounds, we obtain an optimal solution for the problem which is based on vector quantization. Two suboptimal and more practical schemes are also presented as substitutes for the vector quantization scheme. In particular, it is shown that the performance of one of the practical schemes which is called per-symbol quantization is very close to the optimal one. Schemes provided for the inner-product calculations are incorporated into our proposed distributed learning methods for GPs. Experimental results show that with spending few bits per symbol in our communication scheme, our proposed methods outperform previous zero rate distributed GP learning schemes such as Bayesian Committee Model (BCM) and Product of experts (PoE).

A central machine is interested in estimating the underlying structure of a sparse Gaussian Graphical Model (GGM) from a dataset distributed across multiple local machines. The local machines can communicate with the central machine through a wireless multiple access channel. In this paper, we are interested in designing effective strategies where reliable learning is feasible under power and bandwidth limitations. Two approaches are proposed: Signs and Uncoded methods. In the Signs method, the local machines quantize their data into binary vectors and an optimal channel coding scheme is used to reliably send the vectors to the central machine where the structure is learned from the received data. In the Uncoded method, data symbols are scaled and transmitted through the channel. The central machine uses the received noisy symbols to recover the structure. Theoretical results show that both methods can recover the structure with high probability for a large enough sample size. Experimental results indicate the superiority of the Signs method over the Uncoded method under several circumstances.

The problem of motion detection has received considerable attention due to the explosive growth of its applications in video analysis and surveillance systems. While the previous approaches can produce good results, the accurate detection of motion remains a challenging task due to the difficulties raised by illumination variations, occlusion, camouflage, sudden motions appearing in burst, dynamic texture, and environmental changes such as weather conditions, sunlight changes during a day, and so on. In this paper, a novel per-pixel motion descriptor is proposed for motion detection in video sequences which outperforms the current methods in the literature particularly in severe scenarios. The proposed descriptor is based on two complementary three-dimensional discrete wavelet transforms (3D-DWT) and a 3D wavelet leader. In this approach, a feature vector is extracted for each pixel by applying a novel 3D wavelet-based motion descriptor. Then, the extracted features are clustered by the well-known K-means algorithm. The experimental results demonstrate the effectiveness of the proposed method compared to the state-of-the-art approaches in several public benchmark datasets. The application of the proposed method and additional experimental results for several challenging datasets are available online.

Proposing a proper method for face recognition is still a challenging subject in biometric and computer vision applications. Although some reliable systems were introduced under relatively controlled conditions, their recognition rate is not satisfactory in the general settings. This is especially true when there are variations in pose, illumination, and facial expression. To alleviate these problems, a hybrid face recognition system is proposed which benefits from the superiority of both convolutional neural network (CNN) and support vector machine (SVM). To this end, first a genetic algorithm is employed to find the optimum structure of CNN. Then, the performance of the system is improved by replacing the last layer of CNN with an ensemble of SVMs. Finally, using concepts of error correction, decision is made. The potential of CNN as a trainable feature extractor provides a flexible recognition system that can recognise faces with variations in pose and illumination. Simulation results show that the system achieves good recognition rate and is robust against variations in terms of facial expressions, occlusion, noise, and illuminations.

To evaluate the performance of distributed medium access control (MAC) layer of emerging ultra wideband (UWB) and 60-GHz millimeter wave (mmWave) wireless personal area networks (WPANs) based on distributed reservation protocol (DRP) of ECMA-368 and ECMA-387 standards, we provide an analytical model for the probability distribution of steady state throughput as a function of the network dimensions size, number of devices, antenna beamwidth, transmission range, bit error rate (BER), antenna disturbance rate (ADR), aggregation length, payload size, and incoming frame rate. Using this model, the optimal payload size is computed as an example of optimization problem. We also propose a formula forthe starvation and fairness and using them, we found a major flaw in DRP, i.e., its poor fairness. We develop a fair DRP (FDRP) protocol and modify our model to cover it. The models are evaluated by simulation of ad-hoc network scenarios and results show that the average error is less than 2 percent.

To evaluate the performance of the distributed medium access control layer of the emerging ultrawideband and 60-GHz millimeter wave (mmWave) wireless personal area networks based on ECMA-368 and ECMA-387 standards, the first step is to determine the beacon period length (BPL) of the superframe in a given network. In this paper, we provide an analytical model for the probability mass function (PMF) of the BPL as a function of the network dimensions, number of beaconing devices, antenna beamwidth, and the transmission range of the devices. To enable devices with steerable directional antennas in the ECMA-387 standard to have simultaneous communications with neighbors in their different antenna sectors, we propose an improvement to the standard for which we computed the PMF of the BPL in its worst case. The effect of beacon period (BP) contraction on the PMF is also considered and modeled. The proposed model for all cases is evaluated by simulating different scenarios in the network and the results show that on average, the model for the average BPL has an error of 1.2 and 2.5 percent in the current definition of the standard and in the proposed modification, respectively, without BP contraction and 0.9 and 1.5 percent, respectively, with BP contraction.

In this paper, we address the problem of data description using a Bayesian framework. The goal of data description is to draw a boundary around objects of a certain class of interest to discriminate that class from the rest of the feature space. Data description is also known as one-class learning and has a wide range of applications. The proposed approach uses a Bayesian framework to precisely compute the class boundary and therefore can utilize domain information in form of prior knowledge in the framework. It can also operate in the kernel space and therefore recognize arbitrary boundary shapes. Moreover, the proposed method can utilize unlabeled data in order to improve accuracy of discrimination. We evaluate our method using various real-world datasets and compare it with other state of the art approaches of data description. Experiments show promising results and improved performance over other data description and one-class learning algorithms.

Computer-Aided Diagnosis (CAD) systems are widely used for detection of various kinds of abnormalities in mammography images. Masses are one type of these abnormalities which are mostly characterized by their margin and shape. For classification of masses proper features are needed to be extracted. However, the number of well-known features for describing margin is much fewer than geometrical, shape, and textural ones. In addition, most of the existing margin features are highly dependent on segmentation accuracy. In this work, new features for describing margin of masses are presented which can handle inaccuracies in segmentation. These features are obtained from a set of waveforms by wavelet analysis among each of them. For each of these waveforms an edge probability distribution is computed. Then, features are extracted from these probability distributions. Although these features are called margin features, they are highly related to the texture of the mass. For experimentation DDSM dataset was used and our simulations show the great performance of these features in classification of masses.

There is a growing attention in subspace learning using tensor-based approaches in high dimensional spaces. In this paper we first indicate that these methods suffer from the Heteroscedastic problem and then propose a new approach called Heteroscedastic Multilinear Discriminant Analysis (HMDA). Our method can solve this problem by utilizing the pairwise chernoff distance between every pair of clusters with the same index in different classes. We also show that our method is a general form of Multilinear Discriminant Analysis (MDA) approach. Experimental results on CMU-PIE, AR and AT&T face databases demonstrate that the proposed method always perform better than MDA in term of classification accuracy.

In this paper, a new method for path planning is proposed using a genetic algorithm (GA). Our method has two key advantages over existing GA methods. The first is a novel environment representation which allows a more efficient method for obstacles dilation in comparison to current cell based approaches that have a tradeoff between speed and accuracy. The second is the strategy we use to generate the initial population in order to speed up the convergence rate which is completely novel. Simulation results show that our method can find a near optimal path faster than computational geometry approaches and with more accuracy in smaller number of generations than GA methods.

Steganography is the art of hiding information in a cover media without attracting attention. One of the cover media which can be used for steganography is speech. In this paper, we propose a new speech steganography in wavelet domain. In this method, lifting scheme is used to create perfect reconstruction Int2Int wavelets. The data is hidden in some of the Least Significant Bits (LSB) of detail wavelet coefficients. The LSB bits for hiding are selected with a new adaptive algorithm. This algorithm does not hide information in silent parts, so there is no need for silent detection algorithms. This method has zero error in hiding/unhiding process, while normal wavelet domain LSB has about 0.2 % error in equal hiding capacity. This method is a high capacity steganography method which can hide information up to 20% of the input speech. The Signal-to- Noise Ratio (SNR) and listening tests show that the stegano audio is imperceptible from original audio.

The charge recovery logic families have been designed several years ago not in order to eliminate the side-channel leakage but to reduce the power consumption. However, in this article we present a new charge recovery logic style not only to gain high energy efficiency but also to achieve the resistance against side-channel attacks especially against differential power analysis attacks. Our approach is a modified version of a classical charge recovery logic style namely 2N-2N2P. Simulation results show a significant improvement in DPA-resistance level as well as in power consumption reduction in comparison with 2N-2N2P and other DPA-resistant logic styles.

In the Internet telephony, loss of IP packets causes instantaneous discontinuities in the received speech. In this paper, we have focused on finding an error resilient method for this problem. Our proposed method creates artificial correlation between speech samples that pre-distorts the speech signal. The receiver uses this correlation to reconstruct the lost speech packets. An appropriate speech enhancement technique is designed for the reduction of the processing error in the recovered speech caused by the speech codecs. The SegSNR results show the superiority of our proposed speech enhancement method over a recently proposed one.

According to international technology roadmap for semiconductors (ITRS), before the end of this decade, we will be entering the era of a billion transistors on a single chip. The major threat toward the achievement of billion transistor on a chip is poor scalability of current interconnect infrastructure. With the advent of "network on chip (NoC)" various characters and methodologies of traditional networks were hardly considered on-chip. Failure, power and area are the major concepts that should be considered when migrating from traditional interconnection networks to NoCs. In this paper we study the effects of faulty links and nodes on power and performance of mesh based NoC, Also several routing algorithms have been implemented and simulated using a cycle accurate VHDL model of NoC.

This paper proposes two enhanced cross-diamond-hexagonal search algorithms to solve the motion-estimation problem in video coding. These algorithms differ from each other by their second step search only, and both of them employ cross-shaped pattern in first step. Proposed method is an improvement over CDHS, which eliminates some checking points of CDHS algorithm. Experimental results show that the proposed methods perform faster than the diamond search (DS) and CDHS, whereas similar quality is preserved.