SLM (Scalable Logic Module) is a fine-grained reconfigurable logic developed by Kumamoto University. Its small configuration information size characterizes it, resulting in a smaller area for logic cells. We have been developing an SoC-type FPGA called SLMLET to take advantage of SLM. It keeps multiple sets of configuration data in the memory module inside the chip in a compressed form and exchanges them quickly. This paper proposes a simple run-length compression technique called TLC (Tag Less Compression). It achieved a 1.01-3.06 compression ratio, is embedded in the prototype of the SLMLET, and is available now. Then, we propose DMC (Duplication Module Compression), which uses repeatedly appearing patterns in the SLM configuration data. The DMC achieves a better compression ratio for complicated designs that are hard to compress with TLC.

In this study, we devise several seat selection screens for a movie theater ticket reservation system that applies nudges to achieve spatial crowd smoothing without relying on economic incentives. We design three types of nudges that achieve the following: (i) render seats in less-crowded areas noticeable; (ii) present social norms; and (iii) suggest seats in less-crowded areas to people who have selected seats in crowded areas. Results of verification experiment show that (ii) the presentation of social norms is generally effective in avoiding congestion regardless of the ticket sales and (ii) the text of the presented social norms is more effective in avoiding congestion when it contains motivational sentences than when it is verbally expressed. Furthermore, the results indicate that (i) rendering seats in less-crowded areas more conspicuous and (iii) suggesting seats in less-crowded areas to those who select seats in more crowded areas may be effective in avoiding congestion, depending on the ticket sales. Consequently, the feasibility of spatial crowd smoothing without relying on economic incentives for the seat selection screen of a ticket reservation system that applies nudges is demonstrated.

Sign constraints are a handy representation of domain-specific prior knowledge that can be incorporated to machine learning. This paper presents new stochastic dual coordinate ascent (SDCA) algorithms that find the minimizer of the empirical risk under the sign constraints. Generic surrogate loss functions can be plugged into the proposed algorithm with the strong convergence guarantee inherited from the vanilla SDCA. The prediction performance is demonstrated on the classification task for microbiological water quality analysis.

As one of electrical components in transmission lines, vibration damper plays a role in preventing the power lines dancing, and its recognition is an important task for intelligent inspection. However, due to the complex background interference in aerial images, current deep learning algorithms for vibration damper detection often lack accuracy and robustness. To achieve vibration damper detection more accurately, in this study, improved You Only Look Once (YOLO) model is proposed for performing damper detection. Firstly, a damper dataset containing 1900 samples with different scenarios was created. Secondly, the backbone network of YOLOv4 was improved by combining the Res2Net module and Dense blocks, reducing computational consumption and improving training speed. Then, an improved path aggregation network (PANet) structure was introduced in YOLOv4, combined with top-down and bottom-up feature fusion strategies to achieve feature enhancement. Finally, the proposed YOLO model and comparative model were trained and tested on the damper dataset. The experimental results and analysis indicate that the proposed model is more effective and robust than the comparative models. More importantly, the average precision (AP) of this model can reach 98.8%, which is 6.2% higher than that of original YOLOv4 model; and the prediction speed of this model is 62 frames per second (FPS), which is 5 FPS faster than that of YOLOv4 model.

Image classification is a typical computer vision task widely used in practical applications. The images used for training image classification networks are often clean, i.e., without any image degradation. However, Convolutional neural networks trained on clean images perform poorly on degraded or corrupted images in the real world. In this study, we effectively utilize robust data augmentation (DA) with knowledge distillation to improve the classification performance of degraded images. We first categorize robust data augmentations into geometric-and-color and cut-and-delete DAs. Next, we evaluate the effectual positioning of cut-and-delete DA when we apply knowledge distillation. Moreover, we also experimentally demonstrate that combining the RandAugment and Random Erasing approach for geometric-and-color and cut-and-delete DA improves the generalization of the student network during the knowledge transfer for the classification of degraded images.

Register automaton (RA) is an extension of finite automaton for dealing with data values in an infinite domain. In the previous work, we proposed disjunctive μ^↓-calculus (μ^↓_d-calculus), which is a subclass of modal µ-calculus with the freeze quantifier, and showed that it has the same expressive power as RA. However, μ^↓_d-calculus is defined as a logic on finite words, whereas temporal specifications in model checking are usually given in terms of infinite words. In this paper, we re-define the syntax and semantics of μ^↓_d-calculus to be suitable for infinite words and prove that the obtained temporal logic, called μ^↓_dω-calculus, has the same expressive power as Büchi RA.

With the rise of cyber threats, identifying APT groups becomes increasingly crucial for enterprise security experts. This paper introduces a comprehensive framework for profiling APT groups, focusing on Lazarus and APT29. It underscores the critical role of malware hash unit profiling in contemporary cyber security efforts, aiming to fortify organizational defenses against evolving APT threats.

Accurate water level prediction systems improve safety and quality of life. This study introduces a method that uses clustering and deep learning of multisite data to enhance the water level prediction of the Three Gorges Dam. The results show that Cluster-GRU-based can provide accurate forecasts for up to seven days.

Drug discovery, characterized by its time-consuming and costly nature, demands approximately 9 to 17 years and around two billion dollars for development. Despite the extensive investment, about 90% of drugs entering clinical trials face withdrawal, with compound toxicity accounting for 30% of these instances. Ethical concerns and the discrepancy in mechanisms between humans and animals have prompted regulatory restrictions on traditional animal-based toxicity prediction methods. In response, human pluripotent stem cell-based approaches have emerged as an alternative. This paper investigates the scalability challenges inherent in utilizing pluripotent stem cells due to the costly nature of RNAseq and the lack of standardized protocols. To address these challenges, we propose applying Mixup data augmentation, a successful technique in deep learning, to kernel SVM, facilitated by Stochastic Dual Coordinate Ascent (SDCA). Our novel approach, Exact SDCA, leverages intermediate class labels generated through Mixup, offering advancements in both efficiency and effectiveness over conventional methods. Numerical experiments reveal that Exact SDCA outperforms Approximate SDCA and SGD in attaining optimal solutions with significantly fewer epochs. Real data experiments further demonstrate the efficacy of multiplexing gene networks and applying Mixup in toxicity prediction using pluripotent stem cells.

The relevant model based on convolutional neural networks (CNNs) has been proven to be an effective solution in speech enhancement algorithms. However, there needs to be more research on CNNs based on microphone arrays, especially in exploring the correlation between networks associated with different microphones. In this paper, we proposed a CNN-based feature integration network for speech enhancement in microphone arrays. The input of CNN is composed of short-time Fourier transform (STFT) from different microphones. CNN includes the encoding layer, decoding layer, and skip structure. In addition, the designed feature integration layer enables information exchange between different microphones, and the designed feature fusion layer integrates additional information. The experiment proved the superiority of the designed structure.

Recent deep-learning-based video compression models have demonstrated superior performance over traditional codecs. However, few studies have focused on deep learning rate control. In this paper, end-to-end rate control is proposed for deep contextual video compression (DCVC). With the designed two-branch residual-based network, the optimal bit rate ratio is predicted according to the feature correlation of the adjacent frames. Then, the bit rate can be reasonably allocated for every frame by satisfying the temporal feature. To minimize the rate distortion (RD) cost, the optimal λ of the current frame can be obtained from a two-branch regression-based network using the temporal encoded information. The experimental results show that the achievable BD-rate (PSNR) and BD-rate (SSIM) of the proposed algorithm are -0.84% and -0.35%, respectively, with 2.25% rate control accuracy.

Deep networks are undergoing rapid development. However, as the depth of networks increases, the issue of how to fuse features from different layers becomes increasingly prominent. To address this challenge, we creatively propose a cross-layer feature fusion module based on neural dendrites, termed dendritic learning-based feature fusion (DFF). Compared to other fusion methods, DFF demonstrates superior biological interpretability due to the nonlinear capabilities of dendritic neurons. By integrating the classic ResNet architecture with DFF, we devise the ResNeFt. Benefiting from the unique structure and nonlinear processing capabilities of dendritic neurons, the fused features of ResNeFt exhibit enhanced representational power. Its effectiveness and superiority have been validated on multiple medical datasets.

Recent studies on facial expression recognition mainly employs discrete category labels to represent emotion states. However, current intelligent emotion interaction systems require more diverse and precise emotion representation metrics, which has been proposed as Valence, Arousal, Dominance (VAD) multi-dimensional continuous emotion parameters. But there are still very less datasets and methods for VAD analysis, making it difficult to meet the needs of large-scale and high-precision emotion cognition. In this letter, we build multi-dimensional facial expression recognition method by using multi-task learning to improve recognition performance through exploiting the consistency between dimensional and categorial emotions. The evaluation results show that the multi-task learning approach improves the prediction accuracy for VAD multi-dimensional emotion. Furthermore, it applies the method to academic outcomes prediction which verifies that introducing the VAD multi-dimensional and multi-task facial expression recognition is effective in predicting academic outcomes. The VAD recognition code is publicly available on github.com/YeeHoran/Multi-task-Emotion-Recognition.