arXiv subjects

Han Bi

Publications and source records attributed to Han Bi.

5 recordsLinked to original sources

arXiv2026

Nonplanar qubit with tunable gauge symmetry

Circuit quantum electrodynamics embeds Josephson junction qubits within superconducting cavities, and has emerged as a leading approach to quantum computing and quantum simulation. Despite the many permutations of circuit geometry that have been explored, Josephson connectivities have so far been planar, making them effectively low-dimensional. Here we show that a non-planar qubit -- a $3\times3$ crossbar Josephson array -- gives rise to flux-tunable $\mathbb{Z}_3$ combinatorial gauge symmetry (CGS), potentially enabling spin-liquid behavior when networked into a lattice. The observed excitation spectrum shows excellent agreement with predictions from a neural network trained to generate variational quantum states, demonstrating that we have predictive power over our high-dimensional quantum system. Fine-structure splittings near the CGS point are compatible with weak tunneling or symmetry breaking due to experimental imperfections. We additionally use the superconducting cavity to externally induce symmetry breaking, observing a restoration of symmetry at the CGS point where ground states differ only by a $\mathbb{Z}_3$ phase. This work initiates a general program exploring lattice gauge theories using the toolbox of circuit quantum electrodynamics. More broadly, introducing non-planar Josephson connectivities opens a vast space for experimental and theoretical exploration of structures in almost any imaginable dimensionality and geometry.

quant-ph↗

arXiv2025

VM-BHINet:Vision Mamba Bimanual Hand Interaction Network for 3D Interacting Hand Mesh Recovery From a Single RGB Image

Understanding bimanual hand interactions is essential for realistic 3D pose and shape reconstruction. However, existing methods struggle with occlusions, ambiguous appearances, and computational inefficiencies. To address these challenges, we propose Vision Mamba Bimanual Hand Interaction Network (VM-BHINet), introducing state space models (SSMs) into hand reconstruction to enhance interaction modeling while improving computational efficiency. The core component, Vision Mamba Interaction Feature Extraction Block (VM-IFEBlock), combines SSMs with local and global feature operations, enabling deep understanding of hand interactions. Experiments on the InterHand2.6M dataset show that VM-BHINet reduces Mean per-joint position error (MPJPE) and Mean per-vertex position error (MPVPE) by 2-3%, significantly surpassing state-of-the-art methods.

cs.CV↗

arXiv2024

Optical response of edge modes in time-reversal symmetric topological superconductors

Topological superconductors and Majorana edge modes at their boundaries have been theoretically predicted. However, their experimental observation remains controversial. Recent theoretical studies suggest that chiral Majorana edge modes exhibit distinct spatially-resolved optical conductivity compared to chiral Dirac edge modes. In this work, we investigate the optical conductivity and spatially-resolved optical conductivity induced by Majorana edge modes and Dirac edge modes under time-reversal symmetry and crystalline symmetry. We conduct numerical calculations and analytical calculations with edge effective theory for two-dimensional ${\mathbb Z}_2$ topological insulators, strong topological superconductors, and topological crystalline superconductors. Our results show that even under time-reversal symmetry and crystalline symmetry, Majorana edge modes and Dirac edge modes exhibit different optical responses.

cond-mat.supr-con↗

arXiv2023

Inter-band optical transitions of helical Majorana edge modes in topological superconductors

The search for evidence of Majorana states on the edges of topological superconductors (TSCs) is challenging due to the difficulty of detecting such charge-neutral electronic quasiparticles. Local microwave spectroscopy has been shown to be a possible method to detect propagating Majorana modes, where a spatially focused light beam must be used. Here, we show that helical Majorana modes in TSCs allow inter-band transitions and thus contribute to optical conductivity under a spatially uniform light. The existence of such a signal requires the system to break certain symmetries so that the projection of the charge current operator onto helical Majorana edge states leads to inter-band hybridization terms. The general form of this contribution under a tunable time-reversal breaking field is derived, which is valid in the sub-gap low-frequency regime where the edge energy spectrum is linear, and numerical results are obtained in three TSC models, showing remarkable consistency with the analytical prediction. In comparison, the current operator for normal helical edge states, such as in quantum spin Hall insulators, does not cause inter-band transitions and the related optical conductivity vanishes unless the time-reversal symmetry is broken. Our results may help guide feasible experiments to provide evidence of Majorana edge modes in TSCs.

cond-mat.supr-con↗

arXiv2023

Better NTK Conditioning: A Free Lunch from (ReLU) Nonlinear Activation in Wide Neural Networks

Nonlinear activation functions are widely recognized for enhancing the expressivity of neural networks, which is the primary reason for their widespread implementation. In this work, we focus on ReLU activation and reveal a novel and intriguing property of nonlinear activations. By comparing enabling and disabling the nonlinear activations in the neural network, we demonstrate their specific effects on wide neural networks: (a) better feature separation, i.e., a larger angle separation for similar data in the feature space of model gradient, and (b) better NTK conditioning, i.e., a smaller condition number of neural tangent kernel (NTK). Furthermore, we show that the network depth (i.e., with more nonlinear activation operations) further amplifies these effects; in addition, in the infinite-width-then-depth limit, all data are equally separated with a fixed angle in the model gradient feature space, regardless of how similar they are originally in the input space. Note that, without the nonlinear activation, i.e., in a linear neural network, the data separation remains the same as for the original inputs and NTK condition number is equivalent to the Gram matrix, regardless of the network depth. Due to the close connection between NTK condition number and convergence theories, our results imply that nonlinear activation helps to improve the worst-case convergence rates of gradient based methods.

cs.LG↗