Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Yanqin Liu has a total of 38 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Riccati sub-equation neural networks method for exact solutions of partial differential equations

    by Mengke Yu, Jiawei Wang, Yanqin Liu, Xiaoxue Zhang, Run-Fa Zhang, Libo Feng

    Neural networks, with their high expressiveness in function approximation, have become an important research area for finding exact solutions of partial differential equations (PDEs). This paper employs a neural network-based analytical solutions (NNAS) method, which uses explicit neural networks models as trial functions for PDEs. By substituting the trial functions into the PDEs and solving the resulting algebraic systems symbolically, exact solutions can be obtained. However, this method requires designing specific activation functions for each problem, which often limits the diversity of solutions. To overcome this limitation, this paper extends the Riccati sub-equation neural networks (RSENNs) framework by using different solutions forms of the Riccati equation as activation functions in the first hidden layer, and constructing trial functions through feedforward computation. Compared with the NNAS method, the RSENNs method employs the Riccati solutions family as a unified activation functions library, reducing the time required for the complex derivation of activation functions. To verify the effectiveness of this approach, the method is applied to the Burgers equation, Fokker-Planck equation, and KdV-mKdV equation, obtaining various solutions forms, including generalized interaction solutions, generalized rational function solutions, and generalized hyperbolic function solutions. These solutions correspond to wave interaction, singular behavior, and localized structures in nonlinear systems, demonstrating the advantages of the method in solutions diversity. Compared with the NNAS method, the RSENNs method also reduces the time required to find suitable activation functions and improves the efficiency of solving PDEs. This work represents an extension and further development of the RSENNs method, providing an effective approach for finding exact solutions to PDEs based on neural networks.

  • Open Access

    A novel auxiliary equation neural networks method for exactly explicit solutions of nonlinear partial differential equations

    by Shanhao Yuan, Yanqin Liu, Runfa Zhang, Limei Yan, Shunjun Wu, Libo Feng

    In this study, we firstly propose an auxiliary equation neural networks method (AENNM), an innovative analytical method that integrates neural networks (NNs) models with the auxiliary equation method to obtain exact solutions of nonlinear partial differential equations (NLPDEs). A key novelty of this method is the introduction of a novel activation function derived from the solutions of the Riccati equation, establishing a new mathematical link between differential equations theory and deep learning. By combining the strong approximation capability of NNs with the high precision of symbolic computation, AENNM significantly enhances computational efficiency and accuracy. To demonstrate the effectiveness of the AENNM in solving NLPDEs, three numerical examples are investigated, including the nonlinear evolution equation, the Korteweg-de Vries-Burgers equation, and the (2+1)-dimensional Boussinesq equation. Furthermore, some new trial functions are constructed by setting specific activation functions within the "2-2-2-1" and "3-2-2-1" NNs models. By embedding the auxiliary equation method into the NNs framework, we derive previously unreported solutions. The exact analytical solutions are expressed in terms of hyperbolic functions, trigonometric functions, and rational functions. Finally, three-dimensional plots, contour plots, and density plots are presented to illustrate the dynamic characteristics of the obtained solutions. This research provides a novel methodological framework for addressing NLPDEs, with broad applicability across scientific and engineering fields.


  • Open Access

    Neural Networks-Based Analytical Solver for Exact Solutions of Fractional Partial Differential Equations

    by Shanhao Yuan, Yanqin Liu, Limei Yan, Runfa Zhang, Shunjun Wu

    This paper introduces an innovative artificial neural networks-based analytical solver for fractional partial differential equations (fPDEs), combining neural networks (NNs) with symbolic computation. Leveraging the powerful function approximation ability of NNs and the exactness of symbolic methods, our approach achieves notable improvements in both computational speed and solution precision. The efficacy of the proposed method is validated through four numerical examples, with results visualized using three-dimensional surface plots, contour mappings, and density distributions. Numerical experiments demonstrate that the proposed framework successfully derives exact solutions for fPDEs without relying on data samples. This research provides a novel methodological framework for solving fPDEs, with broad applicability across scientific and engineering fields.


  • Top Cited Papers

    Fractional sub-equation neural networks (fSENNs) method for exact solutions of space–time fractional partial differential equations

    by Jiawei Wang, Yanqin Liu, Limei Yan, Kunling Han, Libo Feng, Runfa Zhang

    Analytical solutions of space–time fractional partial differential equations (fPDEs) are crucial for understanding dynamics features in complex systems and their applications. In this paper, fractional sub-equation neural networks (fSENNs) are first proposed to construct exact solutions of space–time fPDEs. The fSENNs embed the solutions of the fractional Riccati equation into neural networks (NNs). The NNs are a multi-layer computational models that are composed of weights and activation functions between neurons in the input, hidden, and output layers. In fSENNs, every neuron of the first hidden layer is assigned to the solutions of the fractional Riccati equation. In this way, the new trial functions are obtained. The exact solutions of space–time fPDEs can be obtained by fSENNs. In order to verify the rationality of this method, space–time fractional telegraph equation, space–time fractional Fisher equation, and space–time fractional CKdV–mKdV equation are investigated, and generalized fractional hyperbolic function solutions, generalized fractional trigonometric function solutions, and generalized fractional rational solutions are obtained. Since the fractional sub-equation is applied to the NNs model for the first time, more and new solutions can be obtained in this paper. The dynamic characteristics of some solutions corresponding to waves have been demonstrated through some diagrams.


  • Open Access

    Gradient-enhanced fractional physics-informed neural networks for solving forward and inverse problems of the multiterm time-fractional Burger-type equation

    by Shanhao Yuan, Yanqin Liu, Yibin Xu, Qiuping Li, Chao Guo, Yanfeng Shen

    In this paper, we introduced the gradient-enhanced fractional physics-informed neural networks (gfPINNs) for solving the forward and inverse problems of the multiterm time-fractional Burger-type equation. The gfPINNs leverage gradient information derived from the residual of the fractional partial differential equation and embed the gradient into the loss function. Since the standard chain rule in integer calculus is invalid in fractional calculus, the automatic differentiation of neural networks does not apply to fractional operators. The automatic differentiation for the integer order operators and the finite difference discretization for the fractional operators were used to construct the residual in the loss function. The numerical results demonstrate the effectiveness of gfPINNs in solving the multiterm time-fractional Burger-type equation. By comparing the experimental results of fractional physics-informed neural networks (fPINNs) and gfPINNs, it can be seen that the training performance of gfPINNs is better than fPINNs.


  • Open Access

    A Fast θ Scheme Combined with the Legendre Spectral Method for Solving a Fractional Klein–Gordon Equation

    by Yanan Li, Yibin Xu, Yanqin Liu, Yanfeng Shen

    In the current work, a fast θ scheme combined with the Legendre spectral method was developed for solving a fractional Klein–Gordon equation (FKGE). The numerical scheme was provided by the Legendre spectral method in the spatial direction, and for the temporal direction, a θ scheme of order O(τ2) with a fast algorithm was taken into account. The fast algorithm could decrease the computational cost from O(M2) to O(MlogM), where M denotes the number of time levels. In addition, correction terms could be employed to improve the convergence rate when the solutions have weak regularity. We proved theoretically that the scheme is unconditionally stable and obtained an error estimate. The numerical experiments demonstrated that our numerical scheme is accurate and efficient.


  • Open Access

    A Compact Scheme Combining the Fast Time Stepping Method for Solving 2D Fractional Subdiffusion Equations

    by Yibin Xu, Yanqin Liu, Xiuling Yin, Libo Feng, Zihua Wang

    In this paper, in order to improve the calculation accuracy and efficiency of α-order Caputo fractional derivative (0 < α ≤ 1), we developed a compact scheme combining the fast time stepping method for solving 2D fractional nonlinear subdiffusion equations. In the temporal direction, a time stepping method was applied. It can reach second-order accuracy. In the spatial direction, we utilized the compact difference scheme, which can reach fourth-order accuracy. Some properties of coefficients are given, which are essential for the theoretical analysis. Meanwhile, we rigorously proved the unconditional stability of the proposed scheme and gave the sharp error estimate. To overcome the intensive computation caused by the fractional operators, we combined a fast algorithm, which can reduce the computational complexity from O(N2) to O(Nlog(N)), where N represents the number of time steps. Considering that the solution of the subdiffusion equation is weakly regular in most cases, we added correction terms to ensure that the solution can achieve the optimal convergence accuracy.