arXiv subjects

Pan Zhang

Publications and source records attributed to Pan Zhang.

At least 145 records · Page 8Linked to original sources

arXiv2021

Seven-coordinated Silicon in a SiO2He Compound Formed under the Extreme Conditions of Planetary Interiors

Changes in atomic coordination numbers at high pressures are fundamental to condensed-matter physics because they initiate the emergence of unexpected structures and phenomena. Silicon is capable of forming eight-, nine-, and ten-coordinated structures under compression,in addition to the usual six-coordinated structures. The missing seven-coordinated silicon remains an open question, but here our theoretical study provides evidence for its existence at high pressures. A combination of a crystal-structure prediction method and first-principles calculations allowed prediction of a stable SiO2He compound containing unique SiO7 polyhedrons, which is a configuration unknown in any proposed silica phase. Consequently, seven-coordinated SiO7 is a possible form of silica at high pressures. Further calculations indicate that the SiO2He phase remains energetically stable with a solid character over a wide range of pressures exceeding 607 GPa and temperatures of 0-9000 K, covering the extreme conditions of the core-mantle boundary in super-Earth exoplanets, or even the Solar System's ice giant planets. Our results may provide theoretical guidance for the discovery of other silicides at high pressures, promote the exploration of materials at planetary core-mantle boundaries, and enable planetary models to be refined.

cond-mat.mtrl-sci↗

arXiv2021

Boltzmann machines as two-dimensional tensor networks

Restricted Boltzmann machines (RBM) and deep Boltzmann machines (DBM) are important models in machine learning, and recently found numerous applications in quantum many-body physics. We show that there are fundamental connections between them and tensor networks. In particular, we demonstrate that any RBM and DBM can be exactly represented as a two-dimensional tensor network. This representation gives an understanding of the expressive power of RBM and DBM using entanglement structures of the tensor networks, also provides an efficient tensor network contraction algorithm for the computing partition function of RBM and DBM. Using numerical experiments, we demonstrate that the proposed algorithm is much more accurate than the state-of-the-art machine learning methods in estimating the partition function of restricted Boltzmann machines and deep Boltzmann machines, and have potential applications in training deep Boltzmann machines for general machine learning tasks.

cond-mat.stat-mech↗

arXiv2021

CoCosNet v2: Full-Resolution Correspondence Learning for Image Translation

We present the full-resolution correspondence learning for cross-domain images, which aids image translation. We adopt a hierarchical strategy that uses the correspondence from coarse level to guide the fine levels. At each hierarchy, the correspondence can be efficiently computed via PatchMatch that iteratively leverages the matchings from the neighborhood. Within each PatchMatch iteration, the ConvGRU module is employed to refine the current correspondence considering not only the matchings of larger context but also the historic estimates. The proposed CoCosNet v2, a GRU-assisted PatchMatch approach, is fully differentiable and highly efficient. When jointly trained with image translation, full-resolution semantic correspondence can be established in an unsupervised manner, which in turn facilitates the exemplar-based image translation. Experiments on diverse translation tasks show that CoCosNet v2 performs considerably better than state-of-the-art literature on producing high-resolution images.

cs.CV↗

arXiv2021

Minimum Long-Loop Feedback Vertex Set and Network Dismantling

Network dismantling aims at breaking a network into disconnected components, and attacking vertices that intersect with many loops has proven to be a most efficient strategy. But the existing loop-focusing methods treat the short loops within densely connected local clusters (e.g., cliques) as equally important as the long loops connecting different clusters. Here we propose for highly clustered artificial and real-world networks a bipartite factor-graph formulation that retains all the long loops while simplifies the local dense clusters as individual factor nodes. We develop a mean-field theory for the associated long-loop feedback vertex set problem and apply its message-passing equation as a solver for network dismantling. The proposed factor-graph loop algorithm outperforms the current state-of-the-art graph loop algorithms by a considerable margin on various real networks. Further improvement in dismantling performance is achievable by optimizing the choice of the local dense clusters.

cs.SI↗

arXiv2021

Simulating the Sycamore quantum supremacy circuits

We propose a general tensor network method for simulating quantum circuits. The method is massively more efficient in computing a large number of correlated bitstring amplitudes and probabilities than existing methods. As an application, we study the sampling problem of Google's Sycamore circuits, which are believed to be beyond the reach of classical supercomputers and have been used to demonstrate quantum supremacy. Using our method, employing a small computational cluster containing 60 graphical processing units (GPUs), we have generated one million correlated bitstrings with some entries fixed, from the Sycamore circuit with 53 qubits and 20 cycles, with linear cross-entropy benchmark (XEB) fidelity equals 0.739, which is much higher than those in Google's quantum supremacy experiments.

quant-ph↗

arXiv2021

Tropical Tensor Network for Ground States of Spin Glasses

We present a unified exact tensor network approach to compute the ground state energy, identify the optimal configuration, and count the number of solutions for spin glasses. The method is based on tensor networks with the Tropical Algebra defined on the semiring. Contracting the tropical tensor network gives the ground state energy; differentiating through the tensor network contraction gives the ground state configuration; mixing the tropical algebra and the ordinary algebra counts the ground state degeneracy. The approach brings together the concepts from graphical models, tensor networks, differentiable programming, and quantum circuit simulation, and easily utilizes the computational power of graphical processing units (GPUs). For applications, we compute the exact ground state energy of Ising spin glasses on square lattice up to 1024 spins, on cubic lattice up to 216 spins, and on 3 regular random graphs up to 220 spins, on a single GPU; We obtain exact ground state energy of (+/-)J Ising spin glass on the chimera graph of D-Wave quantum annealer of 512 qubits in less than 100 seconds and investigate the exact value of the residual entropy of (+/-)J spin glasses on the chimera graph; Finally, we investigate ground-state energy and entropy of 3-state Potts glasses on square lattices up to size 18 x 18. Our approach provides baselines and benchmarks for exact algorithms for spin glasses and combinatorial optimization problems, and for evaluating heuristic algorithms and mean-field theories.

cond-mat.stat-mech↗

arXiv2021

Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation

Self-training is a competitive approach in domain adaptive segmentation, which trains the network with the pseudo labels on the target domain. However inevitably, the pseudo labels are noisy and the target features are dispersed due to the discrepancy between source and target domains. In this paper, we rely on representative prototypes, the feature centroids of classes, to address the two issues for unsupervised domain adaptation. In particular, we take one step further and exploit the feature distances from prototypes that provide richer information than mere prototypes. Specifically, we use it to estimate the likelihood of pseudo labels to facilitate online correction in the course of training. Meanwhile, we align the prototypical assignments based on relative feature distances for two different views of the same target, producing a more compact target feature space. Moreover, we find that distilling the already learned knowledge to a self-supervised pretrained model further boosts the performance. Our method shows tremendous performance advantage over state-of-the-art methods. We will make the code publicly available.

cs.CV↗

arXiv2020

Yao.jl: Extensible, Efficient Framework for Quantum Algorithm Design

We introduce Yao, an extensible, efficient open-source framework for quantum algorithm design. Yao features generic and differentiable programming of quantum circuits. It achieves state-of-the-art performance in simulating small to intermediate-sized quantum circuits that are relevant to near-term applications. We introduce the design principles and critical techniques behind Yao. These include the quantum block intermediate representation of quantum circuits, a builtin automatic differentiation engine optimized for reversible computing, and batched quantum registers with GPU acceleration. The extensibility and efficiency of Yao help boost innovation in quantum algorithm design.

quant-ph↗

arXiv2020

Old Photo Restoration via Deep Latent Space Translation

We propose to restore old photos that suffer from severe degradation through a deep learning approach. Unlike conventional restoration tasks that can be solved through supervised learning, the degradation in real photos is complex and the domain gap between synthetic images and real old photos makes the network fail to generalize. Therefore, we propose a novel triplet domain translation network by leveraging real photos along with massive synthetic image pairs. Specifically, we train two variational autoencoders (VAEs) to respectively transform old photos and clean photos into two latent spaces. And the translation between these two latent spaces is learned with synthetic paired data. This translation generalizes well to real photos because the domain gap is closed in the compact latent space. Besides, to address multiple degradations mixed in one old photo, we design a global branch with apartial nonlocal block targeting to the structured defects, such as scratches and dust spots, and a local branch targeting to the unstructured defects, such as noises and blurriness. Two branches are fused in the latent space, leading to improved capability to restore old photos from multiple defects. Furthermore, we apply another face refinement network to recover fine details of faces in the old photos, thus ultimately generating photos with enhanced perceptual quality. With comprehensive experiments, the proposed pipeline demonstrates superior performance over state-of-the-art methods as well as existing commercial tools in terms of visual quality for old photos restoration.

cs.CV↗

arXiv2020

Supervised Learning with Projected Entangled Pair States

Tensor networks, a model that originated from quantum physics, has been gradually generalized as efficient models in machine learning in recent years. However, in order to achieve exact contraction, only tree-like tensor networks such as the matrix product states and tree tensor networks have been considered, even for modeling two-dimensional data such as images. In this work, we construct supervised learning models for images using the projected entangled pair states (PEPS), a two-dimensional tensor network having a similar structure prior to natural images. Our approach first performs a feature map, which transforms the image data to a product state on a grid, then contracts the product state to a PEPS with trainable parameters to predict image labels. The tensor elements of PEPS are trained by minimizing differences between training labels and predicted labels. The proposed model is evaluated on image classifications using the MNIST and the Fashion-MNIST datasets. We show that our model is significantly superior to existing models using tree-like tensor networks. Moreover, using the same input features, our method performs as well as the multilayer perceptron classifier, but with much fewer parameters and is more stable. Our results shed light on potential applications of two-dimensional tensor network models in machine learning.

cs.CV↗

arXiv2020

Contact Area Detector using Cross View Projection Consistency for COVID-19 Projects

The ability to determine what parts of objects and surfaces people touch as they go about their daily lives would be useful in understanding how the COVID-19 virus spreads. To determine whether a person has touched an object or surface using visual data, images, or videos, is a hard problem. Computer vision 3D reconstruction approaches project objects and the human body from the 2D image domain to 3D and perform 3D space intersection directly. However, this solution would not meet the accuracy requirement in applications due to projection error. Another standard approach is to train a neural network to infer touch actions from the collected visual data. This strategy would require significant amounts of training data to generalize over scale and viewpoint variations. A different approach to this problem is to identify whether a person has touched a defined object. In this work, we show that the solution to this problem can be straightforward. Specifically, we show that the contact between an object and a static surface can be identified by projecting the object onto the static surface through two different viewpoints and analyzing their 2D intersection. The object contacts the surface when the projected points are close to each other; we call this cross view projection consistency. Instead of doing 3D scene reconstruction or transfer learning from deep networks, a mapping from the surface in the two camera views to the surface space is the only requirement. For planar space, this mapping is the Homography transformation. This simple method can be easily adapted to real-life applications. In this paper, we apply our method to do office occupancy detection for studying the COVID-19 transmission pattern from an office desk in a meeting room using the contact information.

cs.CV↗

arXiv2020

Contracting Arbitrary Tensor Networks: General Approximate Algorithm and Applications in Graphical Models and Quantum Circuit Simulations

We present a general method for approximately contracting tensor networks with an arbitrary connectivity. This enables us to release the computational power of tensor networks to wide use in inference and learning problems defined on general graphs. We show applications of our algorithm in graphical models, specifically on estimating free energy of spin glasses defined on various of graphs, where our method largely outperforms existing algorithms including the mean-field methods and the recently proposed neural-network-based methods. We further apply our method to the simulation of random quantum circuits, and demonstrate that, with a trade off of negligible truncation errors, our method is able to simulate large quantum circuits that are out of reach of the state-of-the-art simulation methods.

physics.comp-ph↗

arXiv2020

Helium Incorporation Stabilized Direct-gap Silicides

The search of direct-gap Si-based semiconductors is of great interest due to the potential application in many technologically relevant fields. This work examines the incorporation of He as a possible route to form a direct band gap in Si. Structure predictions and first-principles calculations have shown that He reacts with Si at high pressure, to form the stable compounds Si2He and Si3He. Both compounds have host-guest structures consisting of a channel-like Si host framework filled with He guest atoms. The Si frameworks in two compounds could be persisted to ambient pressure after removal of He, forming two pure Si allotropes. Both Si-He compounds and both Si allotropes exhibit direct or quasi-direct band gaps of 0.84-1.34 eV, close to the optimal value (~1.3 eV) for solar cell applications. Analysis shows that Si2He with an electric-dipole-transition allowed band gap possesses higher absorption capacity than diamond cubic Si, which makes it to be a promising candidate material for thin-film solar cell.

cond-mat.mtrl-sci↗

arXiv2020

Solving Statistical Mechanics on Sparse Graphs with Feedback Set Variational Autoregressive Networks

We propose a method for solving statistical mechanics problems defined on sparse graphs. It extracts a small Feedback Vertex Set (FVS) from the sparse graph, converting the sparse system to a much smaller system with many-body and dense interactions with an effective energy on every configuration of the FVS, then learns a variational distribution parameterized using neural networks to approximate the original Boltzmann distribution. The method is able to estimate free energy, compute observables, and generate unbiased samples via direct sampling without auto-correlation. Extensive experiments show that our approach is more accurate than existing approaches for sparse spin glasses. On random graphs and real-world networks, our approach significantly outperforms the standard methods for sparse systems such as the belief-propagation algorithm; on structured sparse systems such as two-dimensional lattices our approach is significantly faster and more accurate than recently proposed variational autoregressive networks using convolution neural networks.

cond-mat.stat-mech↗

arXiv2020

Bringing Old Photos Back to Life

We propose to restore old photos that suffer from severe degradation through a deep learning approach. Unlike conventional restoration tasks that can be solved through supervised learning, the degradation in real photos is complex and the domain gap between synthetic images and real old photos makes the network fail to generalize. Therefore, we propose a novel triplet domain translation network by leveraging real photos along with massive synthetic image pairs. Specifically, we train two variational autoencoders (VAEs) to respectively transform old photos and clean photos into two latent spaces. And the translation between these two latent spaces is learned with synthetic paired data. This translation generalizes well to real photos because the domain gap is closed in the compact latent space. Besides, to address multiple degradations mixed in one old photo, we design a global branch with a partial nonlocal block targeting to the structured defects, such as scratches and dust spots, and a local branch targeting to the unstructured defects, such as noises and blurriness. Two branches are fused in the latent space, leading to improved capability to restore old photos from multiple defects. The proposed method outperforms state-of-the-art methods in terms of visual quality for old photos restoration.

cs.CV↗

arXiv2020

Cross-domain Correspondence Learning for Exemplar-based Image Translation

We present a general framework for exemplar-based image translation, which synthesizes a photo-realistic image from the input in a distinct domain (e.g., semantic segmentation mask, or edge map, or pose keypoints), given an exemplar image. The output has the style (e.g., color, texture) in consistency with the semantically corresponding objects in the exemplar. We propose to jointly learn the crossdomain correspondence and the image translation, where both tasks facilitate each other and thus can be learned with weak supervision. The images from distinct domains are first aligned to an intermediate domain where dense correspondence is established. Then, the network synthesizes images based on the appearance of semantically corresponding patches in the exemplar. We demonstrate the effectiveness of our approach in several image translation tasks. Our method is superior to state-of-the-art methods in terms of image quality significantly, with the image style faithful to the exemplar with semantic consistency. Moreover, we show the utility of our method for several applications

cs.CV↗

arXiv2020

Gradient Flows of Higher Order Yang-Mills-Higgs Functionals

In this paper, we define a family of functionals generalizing the Yang-Mills-Higgs functional on a closed Riemannian manifold. Then we prove the short time existence of the corresponding gradient flow by a gauge fixing technique. The lack of maximal principle for the higher order operator brings us a lot of inconvenience during the estimates for the Higgs field. We observe that the $L^2$-bound of the Higgs field is enough for energy estimates in $4$ dimension, and we show that, provided the order of derivatives, appearing in the higher order Yang-Mills-Higgs functionals, is strictly greater than 1, solutions to the gradient flow do not hit any finite time singularities. As for the Yang-Mills-Higgs $k$-functional with Higgs self-interaction, we show that, provided $\dim(M)<2(k+1)$, the associated gradient flow admits long time existence with smooth initial data. The proof depends on local $L^2$-derivative estimates, energy estimates and blow-up analysis.

math.DG↗

arXiv2019

Solving Quantum Statistical Mechanics with Variational Autoregressive Networks and Quantum Circuits

We extend the ability of unitary quantum circuits by interfacing it with classical autoregressive neural networks. The combined model parametrizes a variational density matrix as a classical mixture of quantum pure states, where the autoregressive network generates bitstring samples as input states to the quantum circuit. We devise an efficient variational algorithm to jointly optimize the classical neural network and the quantum circuit for quantum statistical mechanics problems. One can obtain thermal observables such as the variational free energy, entropy, and specific heat. As a by product, the algorithm also gives access to low energy excitation states. We demonstrate applications to thermal properties and excitation spectra of the quantum Ising model with resources that are feasible on near-term quantum computers.

quant-ph↗