For classification problems, the traditional least squares twin support vector machine (LSTSVM) generates two nonparallel hyperplanes directly by solving two systems of linear equations instead of a pair of quadratic programming problems (QPPs), which makes LSTSVM much faster than the original TSVM. But the standard LSTSVM adopting quadratic loss measured by the minimal distance is sensitive to noise and unstable to re-sampling. To overcome this problem, the expectile distance is taken into consideration to measure the margin between classes and LSTSVM with asymmetric squared loss (aLSTSVM) is proposed. Compared to the original LSTSVM with the quadratic loss, the proposed aLSTSVM not only has comparable computational accuracy, but also performs good properties such as noise insensitivity, scatter minimization and re-sampling stability. Numerical experiments on synthetic datasets, normally distributed clustered (NDC) datasets and University of California, Irvine (UCI) datasets with different noises confirm the great performance and validity of our proposed algorithm.

Knowledge tracking (KT) algorithm, which can model the cognitive level of learners, is a fundamental artificial intelligence approach to solve the personalized learning problem in the field of education. The recently presented separated self-attentive neural knowledge tracing (SAINT) algorithm has got a great improvement on predictingthe accuracy of students' answers in comparison with the present other methods. However there is still potential to enhance its performance for it fails to effectively utilize temporal features. In this paper, an optimization algorithm for SAINT based on Ebbinghaus' law of forgetting was proposed which took temporal features into account. The proposed algorithm used forgetting law-based data binning to discretize the time information sequences, so as to obtain the temporal featuresin accordance with people's forgetting pattern. Then the temporal features were used as input in the decoder of SAINT model to improve its accuracy. Ablation experiments and comparison experiments were performed on the EdNet dataset in order to verify the effectiveness of the proposed model. Seen in the experimental results,it achieved higher area under curve (AUC) values than the other present representative knowledge tracing algorithms. It demonstrates that temporal featuresare necessary for KT algorithms if it can be properly dealt with.

Facial expression recognition (FER) is a vital application of image processing technology. In this paper, a FER model based on the residual network is proposed. The proposed model introduces the idea of the DenseNet, in which the outputs of the residual blocks are not simply added but are linked to the channel dimension. In addition, transfer learning is used to reduce training costs and accelerate training speed. The accuracy and robustness of the proposed FER model were tested by K-fold cross-validation. Experimental results show that the proposed method has competitive performances on FER2013, FER plus (FERPlus), and the real-world affective faces database (RAF-DB).

In this paper, the resource allocation optimization for the simultaneous wireless information and power transfer (SWIPT) full-duplex (FD) relaying networks is investigated, in which the power-constrained relay scavenges energy from the source signal and assists information transmission by FD operation. Taking into account non-linear energy harvesting (EH) hardware circuit characteristics of the relay, the information rate maximization problem is developed by jointly optimizing the time-switching (TS) factor and transmission powers of the source in two different phases. However, the formulated optimization is highly non-convex and difficult to solve. To cope with this problem, the primary problem is decomposed into two sub-problems with respect to the TS factor and transmission powers. After solving these two sub-problems, the final sub-optimal solutions can be obtained by alternating search. Simulation results prove that the optimization of both TS factor and transmission powers can effectively enhance the information rate for the considered networks.

A source enumeration method based on diagonal loading of eigenvalues and constructing second-order statistics is proposed, for the case that the antenna array observed signals are overlapped with spatial colored noise, and the number of antennas compared with the number of snapshots meet the requirement of general asymptotic regime. Firstly, the sample covariance matrix of the observed signals is obtained, the eigenvalues of the sample covariance matrix can be acquired by eigenvalue decomposition, and the eigenvalues are diagonally loaded, and a new formula for calculating the diagonal loading is presented. Based on the diagonal loaded eigenvalues, the difference values are calculated for the adjacent eigenvalues after diagonal loading, and the statistical variance of the difference values is calculated. On this basis, the second-order statistics of the difference values are constructed, and when the second-order statistics are minimized, the corresponding number of sources is estimated. The proposed method has wide applicability, which is suitable for both general asymptotic regime and classical asymptotic system, and is suitable for both white Gaussian noise environment and colored noise environment. The method makes up for the lack of source enumeration methods in the case of general asymptotic system and colored noise.

This paper proposes a robust adaptive filter based on the exponent sin cost to improve the capability against Gaussian or multiple types of non-Gaussian noises of the adaptive filtering algorithm when dealing with time-varying/ time-invariant linear systems function exponent sin (ExpSin). Then a variable step-size (VSS)-ExpSin algorithm is extended further. Besides, the stepsize, the convergence, and the steady-state performance of the proposed algorithm are validated experimentally. The Monte Carlo simulation results of linear system identification illustrate the principle and efficiency of this proposed adaptive filtering algorithm. Results suggest that the proposed adaptive filtering algorithm has superior performance when estimating the unknown linear systems under multiple-types measurement noises.

Sharing of the electronic medical records among different hospitals raises serious concern of the leakage of individual privacy for the adoption of the semi trustworthiness of the medical cloud platform. The tracking and revocation of malicious users have become urgent problems. To solve these problems, this paper proposed a traceable and directly revocable medical data sharing scheme. In the scheme, a unique identity parameter (ID), which was generated and embedded in the private key generation phase by the medical service provider (MSP), is used to identify legal authorized user and trace malicious user. Only when attributes satisfy the access policy and user's ID is not in the revocation list can the user calculate the decryption key. Malicious user can be tracked and directly revoked by using the revocation list. Under the assumption of decision bilinear Diffie-Hellman (DBDH), this paper has proved that the scheme is able to achieve security against chosen-plaintext attack (CPA). The performance analysis demonstrates that the sizes of the public key and private key are shorter, and the time overhead is less than other schemes in the public-private key generation, data encryption and data decryption stages.

The sudden surge of various applications poses great challenges to the computation capability of mobile devices. To address this issue, computation offloading to multi-access edge computing (MEC) was proposed as a promising paradigm. This paper studies partial computation offloading scenario by considering time delay and energy consumption, where the task can be splitted into several blocks and computed both in local devices and MEC, respectively. Since the formulated problem is a nonconvex problem, this paper proposes an ant colony-based algorithm to achieve the suboptimal solution. Specifically, the proposed method first establish a multi-user one-MEC scenario, in which user devices are able to offload some part of the task to MEC server. Then, it develops an ant colony-based algorithm to decide the offloading parts and allocation strategy of MEC resources to minimize system cost. Finally, simulation results show the effectiveness of the proposed algorithm in terms of system cost and demonstrate that it outperforms other existing methods.

In this communication, a frequency-, radiation pattern- and polarization-reconfigurable antenna employing liquid metal is presented. Two crossed dipole antennas are surrounded by four independent reflectors and directors to realize multi-beam switching. The length of dipole arms can be adjusted by extracting the liquid metal from the needle tube to achieve frequency reconfiguration. The polarization can be switched by injecting liquid metal into different dipole microfluidic channels. It is simple in design and has multiple reconfigurable capabilities. An antenna with a relative frequency tuning range of 35.8% extending from 2.43 GHz to 3.49 GHz is fabricated. It also can perform 6 kinds of beam steering over a 360 coverage and switch between two different polarizations. The antenna has potential to employ cognitive radio (CR) and base station in wireless systems.

To solve the problem that the performance of the coverage, interference rate, load balance andweak power in the radio frequency identification (RFID) network planning. This paper proposes an elite opposition-based learning and Levy flight sparrow search algorithm (SSA), which is named elite opposition-based learning and Levy flight SSA (ELSSA). First, the algorithm initializes the population by an elite opposed-based learning strategy to enhance the diversity of the population. Second, Levy flight is introduced into the scrounger's position update formula to solve the situation that the algorithm falls into the local optimal solution. It has a probability that the current position is changed by Levy flight. This method can jump out of the local optimal solution. In the end, the proposed method is compared with particle swarm optimization (PSO) algorithm, grey wolf optimzer (GWO) algorithm and SSA in the multiple simulation tests. The simulated results showed that, under the same number of readers, the average fitness of the ELSSA is improved respectively by 3.36%, 5.67% and 18.45%. By setting the different number of readers, ELSSA uses fewer readers than other algorithms. The conclusion shows that the proposed method can ensure a satisfying coverage by using fewer readers and achieving higher comprehensive performance.