Search

We extend the Yang-Lee theory to quantum phase transitions to show how singularity enters the phase transition points in one-dimensional many-body systems. We primarily focus on the distribution of Yang-Lee zeros and its associated Yang-Lee edge singularity of two prototypical models: the Su-Schrieffer-Heeger (SSH) model and the XXZ spin chain. By taking the zero-temperature limit, we show how the Yang-Lee zeros approach the quantum phase transition points on the complex plane of parameters. To characterize the edge singularity induced by Yang-Lee zeros in quantum phase transition, we introduce the entanglement entropy of the ground state to show the edges of Yang-Lee zeros lead to the entanglement transition. We further show our results are also applicable to the general non-interacting parity-time-symmetric Hamiltonians., Comment: 5 pages, 2 figures+9 pages

Motivated by recent experimental realization of a Bose-Einstein condensate (BEC) of dipolar molecules, we develop superfluid transport theory for a dissipative BEC to show that a weak uniform two-body loss can induce phase rigidity, leading to superfluid transport of bosons. A generalized f-sum rule is shown to hold for a dissipative superfluid as a consequence of weak U(1) symmetry. It is also demonstrated that dissipation enhances the stability of a molecular BEC with dipolar interactions. Possible experimental situations for measuring the superfluid fraction and the spectral function are discussed., Comment: 6 pages, 1 figure+27 pages 2 figures

Symmetries exist abundantly in the loss function of neural networks. We characterize the learning dynamics of stochastic gradient descent (SGD) when exponential symmetries, a broad subclass of continuous symmetries, exist in the loss function. We establish that when gradient noises do not balance, SGD has the tendency to move the model parameters toward a point where noises from different directions are balanced. Here, a special type of fixed point in the constant directions of the loss function emerges as a candidate for solutions for SGD. As the main theoretical result, we prove that every parameter $\theta$ connects without loss function barrier to a unique noise-balanced fixed point $\theta^*$. The theory implies that the balancing of gradient noise can serve as a novel alternative mechanism for relevant phenomena such as progressive sharpening and flattening and can be applied to understand common practical problems such as representation normalization, matrix factorization, warmup, and formation of latent representations., Comment: preprint

The stochastic gradient descent (SGD) algorithm is the algorithm we use to train neural networks. However, it remains poorly understood how the SGD navigates the highly nonlinear and degenerate loss landscape of a neural network. In this work, we prove that the minibatch noise of SGD regularizes the solution towards a balanced solution whenever the loss function contains a rescaling symmetry. Because the difference between a simple diffusion process and SGD dynamics is the most significant when symmetries are present, our theory implies that the loss function symmetries constitute an essential probe of how SGD works. We then apply this result to derive the stationary distribution of stochastic gradient flow for a diagonal linear network with arbitrary depth and width. The stationary distribution exhibits complicated nonlinear phenomena such as phase transitions, broken ergodicity, and fluctuation inversion. These phenomena are shown to exist uniquely in deep networks, implying a fundamental difference between deep and shallow models., Comment: Preprint

Motivated by entanglement protection, our work utilizes a resonance effect to enhance optomechanical entanglement in the coherent-state representation. We propose a filtering model to filter out the significant detuning components between a thermal-mechanical mode and its surrounding heat baths in the weak coupling limit. We reveal that protecting continuous-variable entanglement involves the elimination of degrees of freedom associated with significant detuning components, thereby resisting decoherence. We construct a nonlinear Langevin equation of the filtering model and numerically show that the filtering model doubles the robustness of the stationary maximum optomechanical entanglement to the thermal fluctuation noise and mechanical damping. Furthermore, we generalize these results to an optical cavity array with one oscillating end-mirror to investigate the long-distance optimal optomechanical entanglement transfer. Our study breaks new ground for applying the resonance effect to protect quantum systems from decoherence and advancing the possibilities of large-scale quantum information processing and quantum network construction., Comment: 20 pages, 7 figures

Yang and Lee investigated phase transitions in terms of zeros of partition functions, namely, Yang-Lee zeros [Phys. Rev. 87, 404 (1952); Phys. Rev. 87, 410 (1952)]. We show that the essential singularity in the superconducting gap is directly related to the number of roots of the partition function of a BCS superconductor. Those zeros are found to be distributed on a semicircle in the complex plane of the interaction strength due to the Fermi-surface instability. A renormalization-group analysis shows that the semicircle theorem holds for a generic quantum many-body system with a marginal coupling, in sharp contrast with the Lee-Yang circle theorem for the Ising spin system. This indicates that the geometry of Yang-Lee zeros is directly connected to the Fermi-surface instability. Furthermore, we unveil the nonunitary criticality in BCS superconductivity that emerges at each individual Yang-Lee zero due to exceptional points and presents a universality class distinct from that of the conventional Yang-Lee edge singularity., Comment: 5pages, 1 figure+14 pages, 5 figures

Higher rank symmetry and higher moment conservation have been drawn considerable attention from, e.g., subdiffusive transport to fracton topological order. In this paper, we perform a one-loop renormalization group (RG) analysis and show how these phenomena emerge at low energies. We consider a $d$-dimensional model of interacting bosons of d components. At higher-rank-symmetric points with conserved angular moments, the $a$-th bosons have kinetic energy only along the $x^a$ direction. Therefore, the symmetric points look highly anisotropic and fine-tuned. By studying RG in a wide vicinity of the symmetric points, we find that symmetry-disallowed kinetic terms tend to be irrelevant within the perturbative regime, which potentially leads to emergent higher-rank symmetry and higher-moment conservation at the deep infrared limit. While non-perturbative analysis is called for in the future, by regarding higher-rank symmetry as an emergent phenomenon, the RG analysis presented in this paper holds alternative promise for realizing higher-rank symmetry and higher-moment conservation in experimentally achievable systems.

For a long period of time, we have been seeking how Berry curvature influnces the transport properties in materials breaking time-reversal symmetry. In time-reversal symmetric material, there will be no thermoelectric current induced by Berry curvature in linear regime. However, the nonlinear Hall current can be shown in non-magnetic and non-centrosymmetric materials, where Berry curvature dipole plays an important role. Most studies are developed from semi-classical Boltzmann equation. Here we show the quantum kinetic theory for nonlinear Nernst effect and introduce a new type of Berry curvature dipole: thermoelectric Berry curvature dipole. This new Berry curvature dipole will also induce the thermoelectric transport in nonlinear regime even in time-reversal invariant crystals. We will also apply our theory to topological crystalline insulator with tilted Dirac cone., Comment: 12 pages, 2 figures, submitted

Vehicle Re-ID has recently attracted enthusiastic attention due to its potential applications in smart city and urban surveillance. However, it suffers from large intra-class variation caused by view variations and illumination changes, and inter-class similarity especially for different identities with the similar appearance. To handle these issues, in this paper, we propose a novel deep network architecture, which guided by meaningful attributes including camera views, vehicle types and colors for vehicle Re-ID. In particular, our network is end-to-end trained and contains three subnetworks of deep features embedded by the corresponding attributes (i.e., camera view, vehicle type and vehicle color). Moreover, to overcome the shortcomings of limited vehicle images of different views, we design a view-specified generative adversarial network to generate the multi-view vehicle images. For network training, we annotate the view labels on the VeRi-776 dataset. Note that one can directly adopt the pre-trained view (as well as type and color) subnetwork on the other datasets with only ID information, which demonstrates the generalization of our model. Extensive experiments on the benchmark datasets VeRi-776 and VehicleID suggest that the proposed approach achieves the promising performance and yields to a new state-of-the-art for vehicle Re-ID.

In this work, we report the set-up and results of the Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with the IEEE International Symposium on Biomedical Imaging (ISBI) 2017 and the International Conferences on Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2017 and 2018. The image dataset is diverse and contains primary and secondary tumors with varied sizes and appearances with various lesion-to-background levels (hyper-/hypo-dense), created in collaboration with seven hospitals and research institutions. Seventy-five submitted liver and liver tumor segmentation algorithms were trained on a set of 131 computed tomography (CT) volumes and were tested on 70 unseen test images acquired from different patients. We found that not a single algorithm performed best for both liver and liver tumors in the three events. The best liver segmentation algorithm achieved a Dice score of 0.963, whereas, for tumor segmentation, the best algorithms achieved Dices scores of 0.674 (ISBI 2017), 0.702 (MICCAI 2017), and 0.739 (MICCAI 2018). Retrospectively, we performed additional analysis on liver tumor detection and revealed that not all top-performing segmentation algorithms worked well for tumor detection. The best liver tumor detection method achieved a lesion-wise recall of 0.458 (ISBI 2017), 0.515 (MICCAI 2017), and 0.554 (MICCAI 2018), indicating the need for further research. LiTS remains an active benchmark and resource for research, e.g., contributing the liver-related segmentation tasks in \url{http://medicaldecathlon.com/}. In addition, both data and online evaluation are accessible via \url{www.lits-challenge.com}., Comment: Patrick Bilic, Patrick Christ, Hongwei Bran Li, and Eugene Vorontsov made equal contributions to this work. Published in Medical Image Analysis