In cellular networks, cell handover refers to the process where a device switches from one base station to another, and this mechanism is crucial for balancing the load among different cells. Traditionally, engineers would manually adjust parameters based on experience. However, the explosive growth in the number of cells has rendered manual tuning impractical. Existing research tends to overlook critical engineering details in order to simplify handover problems. In this paper, we classify cell handover into three types, and jointly model their mutual influence. To achieve load balancing, we propose a multi-agent-reinforcement-learning (MARL)-based scheme to automatically optimize the parameters. Experimental results show that our proposed scheme outperforms existing benchmarks in balancing load and improving network performance.

A common characteristic in integer linear programs (ILPs) is symmetry, allowing variables to be permuted without altering the underlying problem structure. Recently, GNNs have emerged as a promising approach for solving ILPs. However, a significant challenge arises when applying GNNs to ILPs with symmetry: classic GNN architectures struggle to differentiate between symmetric variables, which limits their predictive accuracy. In this work, we investigate the properties of permutation equivalence and invariance in GNNs, particularly in relation to the inherent symmetry of ILP formulations. We reveal that the interaction between these two factors contributes to the difficulty of distinguishing between symmetric variables. To address this challenge, we explore the potential of feature augmentation and propose several guiding principles for constructing augmented features. Building on these principles, we develop an orbit-based augmentation scheme that first groups symmetric variables and then samples augmented features for each group from a discrete uniform distribution. Empirical results demonstrate that our proposed approach significantly enhances both training efficiency and predictive performance.

The combination of linear transformations and nonlinear activation functions forms the foundation of most modern deep neural networks, enabling them to approximate highly complex functions. This paper explores the introduction of quadratic transformations to further increase the nonlinearity of the model, with the aim of enhancing the performance of existing architectures. To minimize the additional parameters and computational burden, we propose a lightweight quadratic enhancer that leverages matrix decomposition, weight sharing, and sparsification techniques. This approach introduces only a minimal and negligible increase in parameters and forward computation, while still yielding substantial improvements in model performance. We evaluate the effectiveness of the proposed method across three tasks: text classification, image classification, and fine-tuning large language models (LLMs). In all tasks, our approach demonstrates significant performance gains.

Precise modeling of the attenuation parameter Q is important to confirm the robust and diagnostic attenuation characteristics of seismic waveforms in oil and gas reservoirs. For this purpose, the nearly constant Q viscoacoustic wave equation is beneficial. Compared with attenuation models such as standard linear solid (SLS), the wave equation corresponding to the nearly constant Q model contains an explicit Q, which is conducive to inversion and can provide a more effective parameterization. This parameterization facilitates the initial suppression of parameter crosstalk in multiparameter inversion. We derive the adjoint equation containing auxiliary variables, compare the sensitive kernels of different parameterizations, and explain the superiority of the constructed parameterization. The truncated Gauss-Newton (GN-TRN) method is introduced to suppress parameter crosstalk further. The GN-TRN method updates the model parameters by calculating the Hessian vector product and iteratively solving the approximation of the Newton gradient directions. The test results of the theoretical model and field data verify the effectiveness and stability of the inversion method.

Although full waveform inversion (FWI) is widely recognized as one of the state-of-the-art techniques in geophysical exploration, there remain several aspects of FWI that require further improvements, specifically in resolution and modeling efficiency. To address this, we introduce an inverse scattering approach to frequency-domain seismic FWI by utilizing a regularized objective function. Different from traditional adjoint methods, the scattering theory allows us to derive the sensitivity kernel explicitly through two Greens’ functions and transforms the nonlinear inverse scattering problem into a series of linear inverse scattering problems, thereby facilitating the calculation of the gradient and Hessian. To mitigate the computational cost when calculating the background and actual wavefields, the fast Fourier transform (FFT) combined with the Krylov subspace method is used to solve the Lippmann-Schwinger (L-S) integral equation (IE) iteratively. Additionally, we incorporate minimum support (MS) stabilizing functional as an extra model misfit term alongside the traditional data misfit function, for a better recovery of the shape structure within the model. Furthermore, the inversion framework is enhanced by integrating an improved limited memory Broyden-Fletcher–Goldfarb-Shanno algorithm, with the regularized Hessian serving as a preconditioner. To demonstrate the efficacy of our method, numerical tests on Marmousi and BP models are presented to validate the performance and robustness of the proposed approach.

Integer linear programs (ILPs) are commonly employed to model diverse practical problems such as scheduling and planning. Recently, machine learning techniques have been utilized to solve ILPs. A straightforward idea is to train a model via supervised learning, with an ILP as the input and an optimal solution as the label. An ILP is symmetric if its variables can be permuted without changing the problem structure, resulting in numerous equivalent and optimal solutions. Randomly selecting an optimal solution as the label can introduce variability in the training data, which may hinder the model from learning stable patterns. In this work, we incorporate the intrinsic symmetry of ILPs and propose a novel training framework called SymILO. Specifically, we modify the learning task by introducing solution permutation along with neural network weights as learnable parameters and then design an alternating algorithm to jointly optimize the loss function. We conduct extensive experiments on ILPs involving different symmetries and the computational results demonstrate that our symmetry-aware approach significantly outperforms three existing methods----achieving $50.3\\%$, $66.5\\%$, and $45.4\\%$ average improvements, respectively.

Generating face from human eyes, named eyes-to-face generation, is an interesting research topic of face synthesis, which has great potential in the field of public security. One of the main challenges in eyes-to-face generation is the unbalanced information between inputs and outputs, where the outputs are complete facial images while the inputs only contain limited information in the region of eyes. The existing methods generate faces directly from eyes without considering the possibly available facial information (e.g. facial attributes), resulting in inaccurate predictions and high uncertainty in those features less correlated with eyes (e.g. hairstyle, moustache, facial contour). To address this challenge, we propose a two-stage solution (named EA2F-GAN) to dynamically optimize eyes-to-face generation via attribute vocabulary. In addition, a dataset named TEAF is constructed based on the public datasets CelebA and LFW, containing 138,934 triples of eye image, attribute vocabulary, and face image. Sufficient experimental results show that, by incorporating additional facial attributes, our proposed approach can synthesize realistic face with high consistency to the original one, significantly overwhelming state-of-the-art methods.

Parkinson's disease is a commonly encountered central neurodegenerative disease in elder people. In recent years, considerable progresses have been made in the etiopathogenesis and treatment of Parkinson's disease. But still, it cannot be prevented. No radical treatments have been found and problems like adverse reactions have to be solved. According to Jueyin of six-meridian syndrome differentiation and the clinical manifestations of Parkinson's disease, Professor Xiaodong Luo exploits differentiations of tremor-spasm and yin-yang, therefore creates effective basic prescriptions for Parkinson's disease.

This paper summarizes a methodology for reliability prediction of new products where field data are sparse, and the allowed number & length of experiments are limited. The methodology relies on estimating a set where the unknown parameters are most likely to be found, calculation of an upper bound for the reliability metric of interest conditioned that the parameters reside in the estimated set, and tightening the bounds via design of experiments. Models of failure propagation, failure acceleration, system operations, and time/cycle to failure at various levels of fidelity & expert elicited information may be incorporated to enhance the accuracy of the predictions. The application of the model is illustrated through numerical studies.

Since exact maximum likelihood (ML) detection is computationally intractable in general, approximate ML approaches are needed to reduce the computation time while maintaining low bit error rate (BER). In this work, we develop an efficient approximate ML decoder for constant modulus signals based on a simple nonlinear programming relaxation. Unlike the existing sphere decoder whose expected complexity is cubic in problem size and whose performance deteriorates with increasing problem size and noise level, our proposed new decoder enjoys a worst case quadratic complexity and scales gracefully with problem dimension and noise level. Our initial testing and analysis suggests that this new decoder is capable of delivering ML like BER performance for PSK signals while requiring substantially lower computational complexity. In this sense, our new decoder is similar to the sphere decoder which is an effective method for QAM signals.