Implicit Neural Representations (INRs) have become a powerful paradigm for modeling continuous signals in computer vision, graphics, and scientific computing. However, multilayer perceptrons (MLPs) generally suffer from severe spectral bias, which limits their ability to accurately model high-frequency details and multi-scale structures. To address this challenge, we propose a novel Multi-Scale Sine Activation (MSA), which explicitly introduces multi-scale frequency responses by incorporating multiple sets of sine activations with logarithmically spaced frequencies in parallel at each layer. MSA is further combined with an amplitude modulation mechanism to ensure numerical stability and robust optimization across different frequency channels. We conduct extensive experiments on a series of challenging tasks, including 1D multi-scale function fitting, image representation, video representation, 3D shape representation, and PDEs solving. Experimental results show that MSA outperforms existing state-of-the-art methods in terms of reconstruction accuracy, detail preservation, and training stability.

The development of high-quality datasets is crucial for benchmarking and advancing research in Graphical User Interface (GUI) agents. Despite their importance, existing datasets are often constructed under idealized conditions, overlooking the diverse anomalies frequently encountered in real-world deployments. To address this limitation, we introduce GUI-Robust, a novel dataset designed for comprehensive GUI agent evaluation, explicitly incorporating seven common types of anomalies observed in everyday GUI interactions. Furthermore, we propose a semi-automated dataset construction paradigm that collects user action sequences from natural interactions via RPA tools and then generate corresponding step and task descriptions for these actions with the assistance of MLLMs. This paradigm significantly reduces annotation time cost by a factor of over 19 times. Finally, we assess state-of-the-art GUI agents using the GUI-Robust dataset, revealing their substantial performance degradation in abnormal scenarios. We anticipate that our work will highlight the importance of robustness in GUI agents and inspires more future research in this direction. The dataset and code are available at https://github.com/chessbean1/GUI-Robust.

The quest for analytical solutions to differential equations has traditionally been constrained by the need for extensive mathematical expertise. Machine learning methods like genetic algorithms have shown promise in this domain, but are hindered by significant computational time and the complexity of their derived solutions. This paper introduces **SSDE** (Symbolic Solver for Differential Equations), a novel reinforcement learning-based approach that derives symbolic closed-form solutions for various differential equations. Evaluations across a diverse set of ordinary and partial differential equations demonstrate that SSDE outperforms existing machine learning methods, delivering superior accuracy and efficiency in obtaining analytical solutions.

Symbolic regression (SR) is the process of finding an unknown mathematical expression given the input and output and has important applications in interpretable machine learning and knowledge discovery. The major difficulty of SR is that finding the expression structure is an NP-hard problem, which makes the entire process time-consuming. In this study, the solution of expression structures was regarded as a classification problem and solved by supervised learning such that SR can be solved quickly by using the solving experience. Techniques for classification tasks, such as equivalent label merging and sample balance, were used to enhance the robustness of the algorithm. We proposed a symbolic network called DeepSymNet to represent symbolic expressions to improve the performance of the algorithm. DeepSymNet has been proven to have a strong representation ability with a shorter label compared to the current popular representation methods, reducing the search space when predicting. Moreover, DeepSymNet conveniently decomposes SR into two smaller subproblems, which makes solving the problem easier. The proposed algorithm was tested on artificially generated expressions and public datasets and compared with other algorithms. The results demonstrate the effectiveness of the proposed algorithm.

Symbolic regression (SR) is a powerful technique for discovering the underlying mathematical expressions from observed data. Inspired by the success of deep learning, recent deep generative SR methods have shown promising results. However, these methods face difficulties in processing high-dimensional problems and learning constants due to the large search space, and they don’t scale well to unseen problems. In this work, we propose DySymNet, a novel neural-guided Dynamic Symbolic Network for SR. Instead of searching for expressions within a large search space, we explore symbolic networks with various structures, guided by reinforcement learning, and optimize them to identify expressions that better-fitting the data. Based on extensive numerical experiments on low-dimensional public standard benchmarks and the well-known SRBench with more variables, DySymNet shows clear superiority over several representative baseline models. Open source code is available at https://github.com/AILWQ/DySymNet.

Head pose classification is an important part of the preprocessing process of face recognition, which can independently solve application problems related to multi-angle. But, due to the impact of the COVID-19 coronavirus pandemic, more and more people wear masks to protect themselves, which covering most areas of the face. This greatly affects the performance of head pose classification. Therefore, this article proposes a method to classify the head pose with wearing a mask. This method focuses on the information that is helpful for head pose classification. First, the H-channel image of the HSV color space is extracted through the conversion of the color space. Then use the line portrait to extract the contour lines of the face, and train the convolutional neural networks to extract features in combination with the grayscale image. Finally, stacked generalization technology is used to fuse the output of the three classifiers to obtain the final classification result. The results on the MAFA dataset show that compared with the current advanced algorithm, the accuracy of our method is 94.14% on the front, 86.58% on the more side, and 90.93% on the side, which has better performance.

Abstract Image quality assessment is to simulate subjective human visual perception and realize image quality inference automatically. Although deep neural networks have achieved great success, the majority of them do not fully consider perception characteristics. Therefore, according to the human visual scale characteristics, we proposed an image quality assessment algorithm based on multiscale and dual domains fusion. Firstly, the original image and its phase congruency respectively input into two branches, feature pyramid and channel attention mechanism are adopted to extract multiscale features. After that, bilinear pool is used to aggregate the spatial and frequency domain characteristics of the corresponding scales, and allows arbitrary scale input to ensure that the features are extracted from the inherent quality images. Finally, the single quality score is obtained through learned weights of each scale. Comparative experiments between our approach and state‐of‐the‐art are conducted on five public databases, the results demonstrate that the proposed algorithm is not only robust to different types and across database, but also sensitive to scale.

Summary In recent years, with the rapid development of adversarial learning technology, facial attribute editing has made great success in a number of areas. Realistic visual effect, invariant identity information, and accurate editing area are the three key issues of facial attribute editing. Unfortunately, most researches focus on the former two problems. However, lack of awareness of the accurate editing area in the task is the main reason for damaging attribute‐irrelevant details. To address this issue, this article proposes a novel facial attribute editing algorithm—a generative adversarial network (GAN) with semantic masks—from the perspective of editing location accuracy. By generating the mask with respect to attribute‐related areas, the semantic segmentation network can only constrain the manipulation in the target region while not harming any attribute‐irrelevant details. The GAN is then combined with the semantic segmentation network to formulate the entire framework, which is referred to as SM‐GAN. Extensive experiments on the public datasets CelebA and LFWA prove that the presented method can not only ensure that the attribute manipulation is realistic, but also allow attribute‐irrelevant regions to remain unchanged. Moreover, it can also simultaneously edit multiple facial attributes.

Summary In recent years, finger vein recognition has attracted more attention and research as a secure method of identification. Convolutional neural networks have achieved great success in the field of finger vein recognition, yet they suffer from high computational complexity, large parameters, and other challenges. To solve these problems, we propose a Gabor convolutional neural network with receptive fields. We use Gabor filters with receptive field properties to design Gabor convolutional layers. Then we replace the conventional convolutional layer with the Gabor convolutional layer; analyze the influence of different loss functions, convolution kernel size, and feature size on the network model; and choose the most suitable model parameters and loss function. Finally, we systematically investigate comparative performance using AGCNN and CNNs in different finger vein databases. Experimental results show that the parameter complexity of AGCNN is significantly less than that of CNNs with a slight performance decrease.

Human pose estimation (HPE) is crucial for computer vision (CV). Moreover, it’s a vital step for computers to understand human actions and behaviours. However, the huge number of parameters and calculations in the HPE model have brought big challenges to deploy to resource-constrained mobile devices. Aiming to overcome the challenge, we propose a sparse pruning method (SPM) for the HPE model. First, L1 regularisation is added in the training phase of the original model, and network parameters of the convolution layers (CLs) and batch normalisation layers (BNLs) are sparsely trained to obtain a network structure with sparse weights. We then combine the sparse weights of filters with the scaling parameters of the BNLs to determine their importance. Finally, the structured pruning method is used to prune the sparse filters and corresponding channels. SPM can reduce the number of model parameters and calculations without affecting precision. Promising results indicate that SPM outperforms other advanced pruning methods.

With the development of deep learning, neural networks are widely used in various fields, and the improved model performance also introduces a considerable number of parameters and computations. Model quantisation is a technique that turns floating-point computing into low-specific-point computing, which can effectively reduce model computation strength, parameter size, and memory consumption but often bring a considerable loss of accuracy. This paper mainly addresses the problem where the distribution of parameters is too concentrated during quantisation aware training (QAT). In the QAT process, we use a piecewise function to statistics the parameter distributions and simulate the effect of quantisation noise in each round of training, based on the statistical results. Experimental results show that by quantising the Transformer network, we lose less precision and significantly reduce the storage cost of the model; compared with the full precision LSTM network, our model has higher accuracy under the condition of a similar storage cost. Meanwhile, compared with other quantisation methods on language modelling task, our approach is more accurate. We validated the effectiveness of our policy on the WikiText-103 and PENN Treebank datasets. The experiments show that our method extremely compresses the storage cost and maintains high model performance.

At present, effectively aggregating and transferring the local features of point cloud is still an unresolved technological conundrum. In this study, we propose a new space-cover convolutional neural network (SC-CNN) for tasks such as point cloud classification and segmentation. The core of this network is space-cover convolution (SC-Conv), which implements depthwise separable convolution on the point cloud. In addition, a newly designed space-cover operator (SCOP) replaces depthwise convolution. The key to SC-Conv is constructing anisotropic spatial geometry in the local point cloud. The SCOP achieves this by utilizing the positional and feature relationships to learn the high-order relationship expression between points. First, data-driven adaptive learning from the 3-D coordinate relationship between the local points is used to determine the weight of the SCOP. Then, the edge feature of the neighboring point relative to the sampling point is used as the input of the SCOP. Finally, a deformable spatial geometry is constructed in the feature space between local points to aggregate the local high-order features. By stacking SC-Conv to construct SC-CNN with a hierarchical network structure for point cloud analysis, we can better perceive the shape information of point cloud and improve network robustness. Finally, we provide numerous experiments to verify that SC-CNN parallels or even outperforms advanced methods in shape classification, part segmentation, and large-scale indoor scene segmentation tasks. The open-source code was published athttps://github.com/changshuowang/SC-CNN.

Modeling, prediction, and recognition tasks depend on the proper representation of the objective curves and surfaces. Polynomial functions have been proved to be a powerful tool for representing curves and surfaces. Until now, various methods have been used for polynomial fitting. With a recent boom in neural networks, researchers have attempted to solve polynomial fitting by using this end-to-end model, which has a powerful fitting ability. However, the current neural network-based methods are poor in stability and slow in convergence speed. In this article, we develop a novel neural network-based method, called Encoder-X, for polynomial fitting, which can solve not only the explicit polynomial fitting but also the implicit polynomial fitting. The method regards polynomial coefficients as the feature value of raw data in a polynomial space expression and therefore polynomial fitting can be achieved by a special autoencoder. The entire model consists of an encoder defined by a neural network and a decoder defined by a polynomial mathematical expression. We input sampling points into an encoder to obtain polynomial coefficients and then input them into a decoder to output the predicted function value. The error between the predicted function value and the true function value can update parameters in the encoder. The results prove that this method is better than the compared methods in terms of stability, convergence, and accuracy. In addition, Encoder-X can be used for solving other mathematical modeling tasks.

Local feature descriptors exhibit great superiority in finger vein recognition due to their stability and robustness against local changes in images. However, most of these are methods use general-purpose descriptors that do not consider finger vein-specific features. In this work, we propose a finger vein-specific local feature descriptors based physiological characteristic of finger vein patterns, i.e., histogram of oriented physiological Gabor responses (HOPGR), for finger vein recognition. First, a prior of directional characteristic of finger vein patterns is obtained in an unsupervised manner. Then the physiological Gabor filter banks are set up based on the prior information to extract the physiological responses and orientation. Finally, to make the feature robust against local changes in images, a histogram is generated as output by dividing the image into non-overlapping cells and overlapping blocks. Extensive experimental results on several databases clearly demonstrate that the proposed method outperforms most current state-of-the-art finger vein recognition methods.