Search

In this paper, a new many-objective optimization evolutionary algorithm (MaOEA), namely Many-Objective Evolutionary Algorithm Based on Spatial Distance and Decision Vector Self-Learning (DVSLEA), is proposed for many-objective optimization. The core idea of the algorithm is to use spatial distance to influence the value of disturbance ratio and then affect the generation of offspring. In order to make the algorithm have a good distribution, distribution vector is introduced for the procedure of disturbance. Moreover, a self-learning process is corporated to ascertain the value of disturbance ratio. To evaluate the performance of DVSLEA, the DTLZ and WFG test suites with 3, 5, 8, 10, and 15 objectives are adopted. The experimental results indicate that DVSLEA shows superior performance over nine competitive evolutionary algorithms(MOEA/DD, NSGA-III, VaEA, SPEA2, SPEA2-SDE, MOEADAWA, onebyoneEA, PREA, RVEA), when solving most of the test problems used. [ABSTRACT FROM AUTHOR]

Multi Image Super-resolution (MISR) refers to the task of enhancing the spatial resolution of a stack of low-resolution (LR) images representing the same scene. Although many deep learning-based single image super-resolution (SISR) technologies have recently been developed, deep learning has not been widely exploited for MISR, even though it can achieve higher reconstruction accuracy because more information can be extracted from the stack of LR images. One of the primary obstacles encountered by deep networks when addressing the MISR problem is the variability in the number of LR images that act as input to the network. This impedes the feasibility of adopting an end-to-end learning approach, because the varying number of input images makes it difficult to construct a training dataset for the network. Another challenge arises from the requirement to align the LR input images to generate high-resolution (HR) image of high quality, which requires complex and sophisticated methods. In this paper, we propose a self-learning based method that can simultaneously perform super-resolution and sub-pixel registration of multiple LR images. The proposed method trains a neural network with only the LR images as input and without any true target HR images; i.e., the proposed method requires no extra training dataset. Therefore, it is easy to use the proposed method to deal with different numbers of input images. To our knowledge this is the first time that a neural network is trained using only LR images to perform a joint MISR and sub-pixel registration. Experimental results confirmed that the HR images generated by the proposed method achieved better results in both quantitative and qualitative evaluations than those generated by other deep learning-based methods. • The proposed joint MISR and sub-pixel registration method uses only the LR images to train the network. • There is no need to construct a dataset to train the network. • The proposed method can be effortlessly adapted to changes in the number of input LR images, without re-training. • The proposed method can be used to generate a dataset of realistically degraded LR images. [ABSTRACT FROM AUTHOR]

Ant Colony Optimization (ACO) algorithms have been widely employed for solving optimization problems. Their ability to find optimal solutions depends heavily on the parameterization of the pheromone trails. However, the pheromone parameterization mechanisms in existing ACO algorithms have two major shortcomings: 1) pheromone trails are instance-specific; hence they need to be generated for each new problem instance, 2) solution is constructed based on static pheromone trails, which ignores the impact of the evolving decisions on the final solution. In this paper, we study the personalized journey route planning problem on multimodal public transport networks (MMPTN) that considers multiple travel criteria. The problem is addressed with a weighted sum method, which provides a journey route that best matches passenger's preference vector consisting of multiple travel criteria. We propose a Machine Learning (ML) based Max–Min Ant System (called ML-MMAS) to solve optimization problems by incorporating ML techniques into Ant Colony Optimization algorithms. ML-MMAS learns a pheromone function to directly produce prominent pheromone trails in order to construct solutions for any new instance, without the need to initialize and update the pheromone trails from scratch. We propose a self-learning framework to train the ML-MMAS using incremental solutions generated by MMAS, hence avoiding the need for pre-computed optimal solutions. Specifically, we develop a deep learning-based pheromone prediction model. We design several groups of features to train the model to characterize the evolving states of the search space during solution construction. Finally, we propose a solution component embedding (SCE) model to learn representations of solution components (transit services), which takes into account the transferability among transit services and passenger transfer preferences. The SCE model enables the extraction of high-quality features for the solution components. It can also be directly applied to solve other optimization problems with solutions that can be modeled as sequences of solution components. We evaluate the proposed ML-MMAS by comparing with exact algorithms and the underlying MMAS, using the MMPTN and passenger demands of Singapore. Results show that ML-MMAS is significantly faster than both the exact algorithm and the original MMAS, while achieving near-optimal solutions. [ABSTRACT FROM AUTHOR]

To ensure the security and stabilization of the communication networks, anomaly detection is the first line of defense. However, their learning process suffers two major issues: (1) inadequate labels : there are many different kinds of attacks but rare abnormal nodes in mt of these atstacks; and (2) inaccurate labels : considering the heavy network flows and new emerging attacks, providing accurate labels for all nodes is very expensive. The inadequate and inaccurate label problem challenges many existing methods because the majority normal nodes result in a biased classifier while the noisy labels will further degrade the performance of the classifier. To tackle these issues, we propose SemiADC, a Semi -supervised A nomaly D etection framework for dynamic C ommunication networks. SemiADC first approximately learns the feature distribution of normal nodes with regularization from abnormal ones. It then cleans the datasets and extracts the nodes sasainaccurate labels by the learned feature distribution and structure-based temporal correlations. These self-learning processes run iteratively with mutual promotion, and finally help increase the accuracy of anomaly detection. Experimental evaluations on real-world datasets demonstrate the effectiveness of our SemiADC, which performs substantially better than the state-of-art anomaly detection approaches without the demand of adequate and accurate supervision. [ABSTRACT FROM AUTHOR]

European schools are atypical international schools, because they do not depend only on an administration, but on institutions and administrations both for their funding and for school curricula. These schools are open to the participation of a number of political and social actors. The libraries of these schools are institutions that reflect the specificity of their schools. The parents' association plays a key role in the library institution. Rather than participating, some parents become involved in funding and in various library activities. This involvement allows parents to better understand their children's new school setting and to strengthen the school-family relation. [ABSTRACT FROM AUTHOR]

• Proposing a binary differential evolution algorithm with self-learning strategy, called MOFS-BDE, to solve multi-objective feature selection problems. • Proposing a new binary mutation operator based on probability difference to guide the individuals to locate potentially optimal areas fast. • Proposing a new one-bit purifying search operator (OPS) for improving the self-learning capability of elite individuals. • Proposing an efficient non-dominated sorting operator with crowding distance to reduce the time consumption of the selection operator in differential evolution. Feature selection is an important data preprocessing method. This paper studies a new multi-objective feature selection approach, called the Binary Differential Evolution with self-learning (MOFS-BDE). Three new operators are proposed and embedded into the MOFS-BDE to improve its performance. The novel binary mutation operator based on probability difference can guide individuals to rapidly locate potentially optimal areas, the developed One-bit Purifying Search operator (OPS) can improve the self-learning capability of the elite individuals located in the optimal areas, and the efficient non-dominated sorting operator with crowding distance can reduce the computational complexity of the selection operator in the differential evolution. Experimental results on a series of public datasets show that the effective combination of the binary mutation and OPS makes our MOFS-BDE achieve a trade-off between local exploitation and global exploration. The proposed method is competitive in comparison with some representative genetic algorithm-, particle swarm-, differential evolution-, and artificial bee colony-based feature selection algorithms. Image, graphical abstract [ABSTRACT FROM AUTHOR]

Prior research has established that dynamically trading-off the performance of the radio-frequency (RF) front-end for reduced power consumption across changing channel conditions, using a feedback control system that modulates circuit and algorithmic level “tuning knobs” in real-time based on received signal quality, leads to significant power savings. It is also known that the optimal power control strategy depends on the process conditions corresponding to the RF devices concerned. This leads to an explosion in the search space needed to find the best feedback control strategy, across all combinations of channel conditions and receiver process corners, making simulation driven optimal control law design impractical and computationally infeasible. Since this problem is largely intractable due to the above complexity of simulation, we propose a self-learning strategy for adaptive RF systems. In this approach, RF devices learn their own performance vs. power consumption vs. tuning knob relationships “on-the-fly” and formulate the most power-optimal control strategy for real-time adaptation of the RF system using neural-network based learning techniques during real-time operation. The methodology is demonstrated using SISO and MIMO RF receiver front-ends as test vehicles and is supported by hardware validation leading to 2.5X-3X power savings with minimal overhead. [ABSTRACT FROM PUBLISHER]