arXiv subjects

Hengxuan Jiang

Publications and source records attributed to Hengxuan Jiang.

3 recordsLinked to original sources

arXiv2025

Self-Attention Assistant Classification of Non-Hermitian Phases in Two-Dimensional Lattice

Classification of the non-Hermitian phases in high-dimensional lattice becomes challenging due to interplay of the band topology and non-Hermiticity. The significant increase in data dimensions and the number of categories has rendered traditional supervised learning and unsupervised manifold learning failed. Here, we propose the self-attention assistant machine learning for clustering non-Hermitian phases in two-dimensional lattice. By incorporating the self-attention mechanism, the model can effectively capture long-range dependencies and important patterns, resulting in a more compact and information-rich latent space. It can achieve Altland-Zirnbauer classification with Bloch vector dataset and distinguish the phases of eigenstates' localized behavior with the competition between non-Hermitian skin effect and topological localization. Our results provide a general method for characterizing non-Hermitian phases in two-dimensional lattice via machine learning.

cond-mat.other↗

arXiv2024

Multiple asymmetric couplings induced unconventional corner mode in topolectrical circuits

We investigate the emergence of unconventional corner mode in a two-dimensional topolectrical circuits induced by asymmetric couplings. The non-Hermitian skin effect of two kinked one-dimensional lattices with multiple asymmetric couplings are explored. Then we extend to the two-dimensional model, derive conditions for the non-Hermitian hybrid skin effect and show how the corner modes are formed by non-reciprocal pumping based on one-dimensional topological modes. We provide explicit electrical circuit setups for realizing our observations via realistic LTspice simulation. Moreover, we show the time varying behaviors of voltage distributions to confirm our results. Our study may help to extend the knowledge on building the topological corner modes in the non-Hermitian presence.

cond-mat.mes-hall↗

arXiv2024

Deep learning for the design of non-Hermitian topolectrical circuits

Non-Hermitian topological phases can produce some remarkable properties, compared with their Hermitian counterpart, such as the breakdown of conventional bulk-boundary correspondence and the non-Hermitian topological edge mode. Here, we introduce several algorithms with multi-layer perceptron (MLP), and convolutional neural network (CNN) in the field of deep learning, to predict the winding of eigenvalues non-Hermitian Hamiltonians. Subsequently, we use the smallest module of the periodic circuit as one unit to construct high-dimensional circuit data features. Further, we use the Dense Convolutional Network (DenseNet), a type of convolutional neural network that utilizes dense connections between layers to design a non-Hermitian topolectrical Chern circuit, as the DenseNet algorithm is more suitable for processing high-dimensional data. Our results demonstrate the effectiveness of the deep learning network in capturing the global topological characteristics of a non-Hermitian system based on training data.

physics.app-ph↗