Explore the vast range of titles available.

Due to the distributed data collection and learning in federated learnings, many clients conduct local training with non-independent and identically distributed (non-IID) datasets. Accordingly, the training from these datasets results in severe performance degradation. We propose an efficient algorithm for enhancing the performance of federated learning by overcoming the negative effects of non-IID datasets. First, the intra-client class imbalance is reduced by rendering the class distribution of clients close to Uniform distribution. Second, the clients to participate in federated learning are selected to make their integrated class distribution close to Uniform distribution for the purpose of mitigating the inter-client class imbalance, which represents the class distribution difference among clients. In addition, the amount of local training data for the selected clients is finely adjusted. Finally, in order to increase the efficiency of federated learning, the batch size and the learning rate of local training for the selected clients are dynamically controlled reflecting the effective size of the local dataset for each client. In the performance evaluation on CIFAR-10 and MNIST datasets, the proposed algorithm achieves 20% higher accuracy than existing federated learning algorithms. Moreover, in achieving this huge accuracy improvement, the proposed algorithm uses less computation and communication resources compared to existing algorithms in terms of the amount of data used and the number of clients joined in the training.

Forest trees are a major source of biogenic volatile organic compounds (BVOCs). Terpenes and terpenoids are known as the main BVOCs of forest aerosols. These compounds have been shown to display a broad range of biological activities in various human disease models, thus implying that forest aerosols containing these compounds may be related to beneficial effects of forest bathing. In this review, we surveyed studies analyzing BVOCs and selected the most abundant 23 terpenes and terpenoids emitted in forested areas of the Northern Hemisphere, which were reported to display anti-inflammatory activities. We categorized anti-inflammatory processes related to the functions of these compounds into six groups and summarized their molecular mechanisms of action. Finally, among the major 23 compounds, we examined the therapeutic potentials of 12 compounds known to be effective against respiratory inflammation, atopic dermatitis, arthritis, and neuroinflammation among various inflammatory diseases. In conclusion, the updated studies support the beneficial effects of forest aerosols and propose their potential use as chemopreventive and therapeutic agents for treating various inflammatory diseases.

In this paper, we propose a method for predicting epileptic seizures using a pre-trained model utilizing supervised contrastive learning and a hybrid model combining residual networks (ResNet) and long short-term memory (LSTM). The proposed training approach encompasses three key phases: pre-processing, pre-training as a pretext task, and training as a downstream task. In the pre-processing phase, the data is transformed into a spectrogram image using short time Fourier transform (STFT), which extracts both time and frequency information. This step compensates for the inherent complexity and irregularity of electroencephalography (EEG) data, which often hampers effective data analysis. During the pre-training phase, augmented data is generated from the original dataset using techniques such as band-stop filtering and temporal cutout. Subsequently, a ResNet model is pre-trained alongside a supervised contrastive loss model, learning the representation of the spectrogram image. In the training phase, a hybrid model is constructed by combining ResNet, initialized with weight values from the pre-trained model, and LSTM. This hybrid model extracts image features and time information to enhance prediction accuracy. The proposed method’s effectiveness is validated using datasets from CHB-MIT and Seoul National University Hospital (SNUH). The method’s generalization ability is confirmed through Leave-one-out cross-validation. From the experimental results measuring accuracy, sensitivity, and false positive rate (FPR), CHB-MIT was 91.90%, 89.64%, 0.058 and SNUH was 83.37%, 79.89%, and 0.131. The experimental results demonstrate that the proposed method outperforms the conventional methods.

This paper presents the design and realization of a compact ultra-wideband (UWB) antenna with a rectangular notch wireless area network (WLAN) band that has controllable notched bandwidth and center frequency. The UWB characteristics of the antenna are achieved by truncating the lower ends of the rectangular microstrip patch, and the notch characteristics are obtained by using electromagnetic bandgap (EBG) structures. EBGs consist of two rectangular metallic conductors loaded on the back of the radiator, which is connected to the patch by shorting pins. A rectangular notch at the WLAN band with high selectivity is realized by tuning the individual resonant frequencies of the EBGs and merging them. Furthermore, the results show that the bandwidth and frequency of the rectangular notch band could be controlled according to the on-demand rejection band applications. In the demonstration, the rectangular notch band was shifted to X-band satellite communication by tuning the EBG parameters. The simulated and measured results show that the proposed antenna has an operational bandwidth from 3.1–12.5 GHz for |S11| < -10 with a rectangular notch band from 5–6 GHz, thus rejecting WLAN band signals. The antenna also has additional advantages: the overall size of the compact antenna is 16 × 25 × 1.52 mm3 and it has stable gain and radiation patterns.

In a cellular communication system, deploying a relay station (RS) is an effective alternative to installing a new base station (BS). A dual-hop network enhances the throughput of mobile stations (MSs) located in shadow areas or at cell edges by installing RSs between BSs and MSs. Because additional radio resources should be allocated to the wireless link between BS and RS, a frame to be transmitted from BS is divided into an access zone (AZ) and a relay zone (RZ). BS and RS communicate with each other through the RZ, and they communicate with their registered MSs through an AZ. However, if too many MSs are registered with a certain BS or RS, MS overloading may cause performance degradation. To prevent such performance degradation, it is very important to find the proper positions for RSs to be deployed. In this paper, we propose a method for finding the sub-optimal RS deployment location for the purpose of load-balancing and throughput enhancement. The advantage of the proposed method is the efficiency in find the sub-optimal location of RSs and its reliable tradeoff between load-balancing throughput enhancement. Since the proposed scheme finds the proper position by adjusting the distance and angle of RSs, its computational complexity lower than other global optimization approach or learning-based approach. In addition, the proposed scheme is constituted with the two stages of load-balancing and throughput enhancement. These procedures result in the appropriate tradeoff between load-balancing and throughput enhancement. The simulation results support these advancements of the proposed scheme.

A network composed of unmanned aerial vehicles (UAVs), serving as base stations (UAV-BS network), is emerging as a promising component in next-generation communication systems. In the UAV-BS network, the optimal positioning of a UAV-BS is an essential requirement to establish line-of-sight (LoS) links for ground users. A novel deep Q-network (DQN)-based learning model enabling the optimal deployment of a UAV-BS is proposed. Moreover, without re-learning of the model and the acquisition of the path information of ground users, the proposed model presents the optimal UAV-BS trajectory while ground users move. Specifically, the proposed model optimizes the trajectory of a UAV-BS by maximizing the mean opinion score (MOS) for ground users who move to various paths. Furthermore, the proposed model is highly practical because, instead of the locations of individual mobile users, an average channel power gain is used as an input parameter. The accuracy of the proposed model is validated by comparing the results of the model with those of a mathematical optimization solver.

In this article, we present a model for analyzing the co-occurrence count data derived from practical fields such as user–item or item–item data from online shopping platforms and co-occurring word–word pairs in sequences of texts. Such data contain important information for developing recommender systems or studying the relevance of items or words from non-numerical sources. Different from traditional regression models, there are no observations for covariates. Additionally, the co-occurrence matrix is typically of such high dimension that it does not fit into a computer’s memory for modeling. We extract numerical data by defining windows of co-occurrence using weighted counts on the continuous scale. Positive probability mass is allowed for zero observations. We present the Shared Parameter Alternating Tweedie (SA-Tweedie) model and an algorithm to estimate the parameters. We introduce a learning rate adjustment used along with the Fisher scoring method in the inner loop to help the algorithm stay on track with optimizing direction. Gradient descent with the Adam update was also considered as an alternative method for the estimation. Simulation studies showed that our algorithm with Fisher scoring and learning rate adjustment outperforms the other two methods. We applied SA-Tweedie to English-language Wikipedia dump data to obtain dense vector representations for WordPiece tokens. The vector representation embeddings were then used in an application of the Named Entity Recognition (NER) task. The SA-Tweedie embeddings significantly outperform GloVe, random, and BERT embeddings in the NER task. A notable strength of the SA-Tweedie embedding is that the number of parameters and training cost for SA-Tweedie are only a tiny fraction of those for BERT.

High-dimensional sparse matrix data frequently arise in various applications. A notable example is the weighted word–word co-occurrence count data, which summarizes the weighted frequency of word pairs appearing within the same context window. This type of data typically contains highly skewed non-negative values with an abundance of zeros. Another example is the co-occurrence of item–item or user–item pairs in e-commerce, which also generates high-dimensional data. The objective is to utilize these data to predict the relevance between items or users. In this paper, we assume that items or users can be represented by unknown dense vectors. The model treats the co-occurrence counts as arising from zero-inflated Gamma random variables and employs cosine similarity between the unknown vectors to summarize item–item relevance. The unknown values are estimated using the shared parameter alternating zero-inflated Gamma regression models (SA-ZIG). Both canonical link and log link models are considered. Two parameter updating schemes are proposed, along with an algorithm to estimate the unknown parameters. Convergence analysis is presented analytically. Numerical studies demonstrate that the SA-ZIG using Fisher scoring without learning rate adjustment may fail to find the maximum likelihood estimate. However, the SA-ZIG with learning rate adjustment performs satisfactorily in our simulation studies.

Time synchronization is an important issue in ad-hoc networks for reliable information exchange. The algorithms for time synchronization in ad-hoc networks are largely categorized into two types. One is based on a selection of a reference node, and the other is based on a consensus among neighbor nodes. These two types of methods are targeting static environments. However, synchronization errors among nodes increase sharply when nodes move or when incorrect synchronization information is exchanged due to the failure of some nodes. In this paper, we propose a synchronization technique for mobile ad-hoc networks, which considers both the mobility of nodes and the abnormal behaviors of malicious or failed nodes. Specifically, synchronization information extracted from a median of the time information of the neighbor nodes is quickly disseminated. This information effectively excludes the outliers, which adversely affect the synchronization of the networks. In addition, Kalman filtering is applied to reduce the synchronization error occurring in the transmission and reception of time information. The simulation results confirm that the proposed scheme has a fast synchronization convergence speed and low synchronization error compared to conventional algorithms.

In this paper, a novel distributed scheduling scheme for an ad-hoc network is proposed. Specifically, the throughput and the delay of packets with different importance are flexibly adjusted by quantifying the importance as weight factors. In this scheme, each node is equipped with two queues, one for packets with high importance and the other for packets with low importance. The proposed scheduling scheme consists of two procedures: intra-node slot reallocation and inter-node reallocation. In the intra-node slot reallocation, self-fairness is adopted as a key metric, which is a composite of the quantified weight factors and traffic loads. This intra-node slot reallocation improves the throughput and the delay performance. Subsequently, through an inter-node reallocation algorithm adopted from LocalVoting (slot exchange among queues having the same importance), the fairness of traffics with the same importance is enhanced. Thorough simulations were conducted under various traffic load and weight factor settings. The simulation results show that the proposed algorithm can adjust packet delivery performance according to a predefined weight factor. Moreover, compared with conventional algorithms, the proposed algorithm achieves better performance in throughput and delay. The low average delay while attaining the high throughput ensures the excellent performance of the proposed algorithm.