arXiv subjects

Xiao Dong

Publications and source records attributed to Xiao Dong.

At least 55 records · Page 3Linked to original sources

arXiv2021

Pinpointing the Memory Behaviors of DNN Training

The training of deep neural networks (DNNs) is usually memory-hungry due to the limited device memory capacity of DNN accelerators. Characterizing the memory behaviors of DNN training is critical to optimize the device memory pressures. In this work, we pinpoint the memory behaviors of each device memory block of GPU during training by instrumenting the memory allocators of the runtime system. Our results show that the memory access patterns of device memory blocks are stable and follow an iterative fashion. These observations are useful for the future optimization of memory-efficient training from the perspective of raw memory access patterns.

cs.PF↗

arXiv2021

Negative linear compressibility and unusual dynamic behaviors of NaB3

First-principles calculations reveal that sodium boride (NaB3) undergoes a phase transition from a tetragonal P4/mbm phase to an orthorhombic Pbam phase at about 16 GPa, accompanied by counterintuitive lattice expansion along the crystallographic a-axis. This unusual compression behavior is identified as negative linear compressibility (NLC), which is dominantly attributed to the symmetry-breaking of boron framework. Meanwhile, the P4/mbm and Pbam phases form superionic conductors after undergoing a peculiar swap state at high temperature. Specifically, under warm conditions the Na cation pairs exhibit a rare local exchange (or rotation) behavior, which may be originated from the asymmetric energy barriers of different diffusion paths. The study of NaB3 compound sheds new light on a material with the combination of NLC and ion transportation at extreme conditions.

cond-mat.mtrl-sci↗

arXiv2021

A Unified Joint Maximum Mean Discrepancy for Domain Adaptation

Domain adaptation has received a lot of attention in recent years, and many algorithms have been proposed with impressive progress. However, it is still not fully explored concerning the joint probability distribution (P(X, Y)) distance for this problem, since its empirical estimation derived from the maximum mean discrepancy (joint maximum mean discrepancy, JMMD) will involve complex tensor-product operator that is hard to manipulate. To solve this issue, this paper theoretically derives a unified form of JMMD that is easy to optimize, and proves that the marginal, class conditional and weighted class conditional probability distribution distances are our special cases with different label kernels, among which the weighted class conditional one not only can realize feature alignment across domains in the category level, but also deal with imbalance dataset using the class prior probabilities. From the revealed unified JMMD, we illustrate that JMMD degrades the feature-label dependence (discriminability) that benefits to classification, and it is sensitive to the label distribution shift when the label kernel is the weighted class conditional one. Therefore, we leverage Hilbert Schmidt independence criterion and propose a novel MMD matrix to promote the dependence, and devise a novel label kernel that is robust to label distribution shift. Finally, we conduct extensive experiments on several cross-domain datasets to demonstrate the validity and effectiveness of the revealed theoretical results.

cs.LG↗

arXiv2020

Accelerating Deep Learning Inference with Cross-Layer Data Reuse on GPUs

Accelerating the deep learning inference is very important for real-time applications. In this paper, we propose a novel method to fuse the layers of convolutional neural networks (CNNs) on Graphics Processing Units (GPUs), which applies data reuse analysis and access optimization in different levels of the memory hierarchy. To achieve the balance between computation and memory access, we explore the fusion opportunities in the CNN computation graph and propose three fusion modes of convolutional neural networks: straight, merge and split. Then, an approach for generating efficient fused code is designed, which goes deeper in multi-level memory usage for cross-layer data reuse. The effectiveness of our method is evaluated with the network layers from state-of-the-art CNNs on two different GPU platforms, NVIDIA TITAN Xp and Tesla P4. The experiments show that the average speedup is 2.02x on representative structures of CNNs, and 1.57x on end-to-end inference of SqueezeNet.

cs.DC↗

arXiv2020

Helium Induced Nitrogen Salt at High Pressure

The energy landscape of helium-nitrogen mixtures is explored by ab initio evolutionary searches, which predicted several stable helium-nitrogen compounds in the pressure range from 25 to 100 GPa. In particular, the monoclinic structure of HeN$_{22}$ consists of neutral He atoms, partially ionic dimers N$_{2}$$^{δ-}$, and lantern-like cages N$_{20}$$^{δ+}$. The presence of helium not only greatly enhances structural diversity of nitrogen solids, but also tremendously lowers the formation pressure of nitrogen salt. The unique nitrogen framework of (HeN$_{20}$)$^{δ+}$N$_{2}$$^{δ-}$ may be quenchable to ambient pressure even after removing helium. The estimated energy density of N$_{20}$$^{δ+}$N$_{2}$$^{δ-}$ (10.44 kJ/g) is $\sim$2.4 times larger than that of trinitrotoluene (TNT), indicating a very promising high-energy-density material.

cond-mat.mtrl-sci↗

arXiv2020

Theoretical study of the pressure-induced structure, phase transition, mechanical and electronic properties in the V-N system

Stable compounds in the V-N system are systematically searched and four new high-pressure phases are found, including C2/m-V$_9$N, Pbam-V$_5$N$_2$, Pnma-V$_2$N and I4/mcm-VN$_2$. V$_2$N undergoes a phase transition from $\varepsilon$-Fe$_2$N-type V$_2$N (P$\bar{3}$1m) to $ζ$-Fe$_2$N-type V$_2$N (Pbcn) at 10 GPa and to Fe$_2$C-type V$_2$N (Pnnm) at 59 GPa, then to Pnma-V$_2$N at 96 GPa. Low-temperature tetragonal VN is theoretically proved to belong to space group P$\bar{4}$2m. The estimated Vickers hardnesses and fracture toughness of WC-type VN are around 37 GPa and 4.3-6.1 MPa m$^{1/2}$, respectively. Al$_2$Cu-type VN$_2$ (I4/mcm) with a Vickers hardness of 25-27 GPa and fracture toughness of 3.6-6.6 MPa m$^{1/2}$ also shows excellent mechanical properties. Elastic properties of WC-type mononitrides of transition metals from IVB group (Ti, Zr and Hf), VB group (V, Nb and Ta) and VIB (Cr, Mo and W) are calculated and compared. Both the bond strength and structural configuration determine the mechanical properties of a material.

cond-mat.mtrl-sci↗

arXiv2019

High-Temperature Superconductivity in the Ti--H System at High Pressures

Search for stable high-pressure compounds in the Ti--H system reveals the existence of titanium hydrides with new stoichiometries, including Ibam-Ti$_2$H$_5$, I4/m-Ti$_5$H$_{13}$, I$\bar{4}$-Ti$_5$H$_{14}$, Fddd-TiH$_4$, Immm-Ti$_2$H$_{13}$, P$\bar{1}$-TiH$_{12}$, and C2/m-TiH$_{22}$. Our calculations predict I4/mmm $\rightarrow$ R$\bar{3}$m and I4/mmm $\rightarrow$ Cmma transitions in TiH and TiH$_2$, respectively. Phonons and the electron--phonon coupling of all searched titanium hydrides are analyzed at high pressure. It is found that Immm-Ti$_2$H$_{13}$ rather than the highest hydrogen content C2/m-TiH$_{22}$, exhibits the highest superconducting critical temperature T$_{c}$. The estimated T$_{c}$ of Immm-Ti$_2$H$_{13}$ and C2/m-TiH$_{22}$ are respectively 127.4--149.4 K ($μ^{*}$=0.1-0.15) at 350 GPa and 91.3--110.2 K at 250 GPa by numerically solving the Eliashberg equations. One of the effects of pressure on T$_{c}$ can be attributed to the softening and hardening of phonons with increasing pressure.

cond-mat.mtrl-sci↗

arXiv2019

Adaptive Collaborative Similarity Learning for Unsupervised Multi-view Feature Selection

In this paper, we investigate the research problem of unsupervised multi-view feature selection. Conventional solutions first simply combine multiple pre-constructed view-specific similarity structures into a collaborative similarity structure, and then perform the subsequent feature selection. These two processes are separate and independent. The collaborative similarity structure remains fixed during feature selection. Further, the simple undirected view combination may adversely reduce the reliability of the ultimate similarity structure for feature selection, as the view-specific similarity structures generally involve noises and outlying entries. To alleviate these problems, we propose an adaptive collaborative similarity learning (ACSL) for multi-view feature selection. We propose to dynamically learn the collaborative similarity structure, and further integrate it with the ultimate feature selection into a unified framework. Moreover, a reasonable rank constraint is devised to adaptively learn an ideal collaborative similarity structure with proper similarity combination weights and desirable neighbor assignment, both of which could positively facilitate the feature selection. An effective solution guaranteed with the proved convergence is derived to iteratively tackle the formulated optimization problem. Experiments demonstrate the superiority of the proposed approach.

cs.IR↗

arXiv2019

Understanding over-parameterized deep networks by geometrization

A complete understanding of the widely used over-parameterized deep networks is a key step for AI. In this work we try to give a geometric picture of over-parameterized deep networks using our geometrization scheme. We show that the Riemannian geometry of network complexity plays a key role in understanding the basic properties of over-parameterizaed deep networks, including the generalization, convergence and parameter sensitivity. We also point out deep networks share lots of similarities with quantum computation systems. This can be regarded as a strong support of our proposal that geometrization is not only the bible for physics, it is also the key idea to understand deep learning systems.

cs.LG↗

arXiv2019

Magnetic borophenes from evolutionary search

A computation methodology based on ab initio evolutionary algorithms and the spin-polarized density functional theory was developed to predict two-dimensional (2D) magnetic materials. Its application to a model system borophene reveals an unexpected rich magnetism and polymorphism. A stable borophene with nonzero thickness was an antiferromagnetic (AFM) semiconductor from first-principles calculations, which can be further turned into a half metal by finite electron doping. In this borophene, the buckling and coupling among three atomic layers are not only responsible for the magnetism, but also result in an out-of-plane negative Poissons ratios under uniaxial tension, making it the first elemental material possessing auxetic and magnetic properties simultaneously.

cond-mat.mtrl-sci↗

arXiv2019

Background subtraction on depth videos with convolutional neural networks

Background subtraction is a significant component of computer vision systems. It is widely used in video surveillance, object tracking, anomaly detection, etc. A new data source for background subtraction appeared as the emergence of low-cost depth sensors like Microsof t Kinect, Asus Xtion PRO, etc. In this paper, we propose a background subtraction approach on depth videos, which is based on convolutional neural networks (CNNs), called BGSNet-D (BackGround Subtraction neural Networks for Depth videos). The method can be used in color unavailable scenarios like poor lighting situations, and can also be applied to combine with existing RGB background subtraction methods. A preprocessing strategy is designed to reduce the influences incurred by noise from depth sensors. The experimental results on the SBM-RGBD dataset show that the proposed method outperforms existing methods on depth data.

cs.CV↗

arXiv2019

Geometrization of deep networks for the interpretability of deep learning systems

How to understand deep learning systems remains an open problem. In this paper we propose that the answer may lie in the geometrization of deep networks. Geometrization is a bridge to connect physics, geometry, deep network and quantum computation and this may result in a new scheme to reveal the rule of the physical world. By comparing the geometry of image matching and deep networks, we show that geometrization of deep networks can be used to understand existing deep learning systems and it may also help to solve the interpretability problem of deep learning systems.

cs.LG↗

arXiv2018

Auto-tuning Neural Network Quantization Framework for Collaborative Inference Between the Cloud and Edge

Recently, deep neural networks (DNNs) have been widely applied in mobile intelligent applications. The inference for the DNNs is usually performed in the cloud. However, it leads to a large overhead of transmitting data via wireless network. In this paper, we demonstrate the advantages of the cloud-edge collaborative inference with quantization. By analyzing the characteristics of layers in DNNs, an auto-tuning neural network quantization framework for collaborative inference is proposed. We study the effectiveness of mixed-precision collaborative inference of state-of-the-art DNNs by using ImageNet dataset. The experimental results show that our framework can generate reasonable network partitions and reduce the storage on mobile devices with trivial loss of accuracy.

cs.DC↗

arXiv2018

Statistical Modeling of an astro-comb for high precision radial velocity observation

The advent of the laser frequency comb as the wavelength calibration unit allows us to measure the radial velocity at $cm\ s^{-1}$ precision level with high stability in long-term, which enable the possibility of the detection of Earth-twins around solar-like stars. Recent study shows that the laser frequency comb can also be used to measure and study the precision of the instrumental system including the variations of line profile and the systematic uncertainty and instrumental drift. In this paper, we present the stringent analysis of a laser frequency comb(LFC) system with 25GHz repetition frequency on a R$\sim$50,000 spectrograph with the wavelength spanning from 5085Å\ to 7380Å. We report a novel fitting model optimized for the comb line profile, the constrained double Gaussian. The constraint condition is set as $\left|μ_{1,2} - μ\right| <\sqrt{2ln2}σ$. We introduce Bayesian information criterion to test various models. Compared to the traditional Gaussian model, the CDG(Constrained Double Gaussians) model provides much better goodness of fit. We apply the CDG model to the observed comb data to demonstrate the improvement of RV precision with CDG model. We find that the improvement of CDG model is about 40\%$\sim$60\% for wavelength calibration precision. We also consider the application to use the LFC and CDG model as a tool to characterize the line shape variation across the detector. The motivation of this work is to measure and understand the details of the comb lines including their asymmetry and behaviors under various conditions, which plays a significant role in the simultaneous calibration process and cross-correlation function method to determine the Doppler shift at high precision level.

astro-ph.IM↗

arXiv2018

The shortness of human life constitutes its limit

In this paper, we affirm our earlier findings of evidence for a limit to human lifespan. In particular, we assess the analyses in extreme value theory (EVT) performed by Rootzén and Zholud. We find that their criticisms of our work are unfounded and that their analyses are contradicted by several other papers using EVT. Furthermore, we find that even if we completely accept the conclusions about late-life human mortality reached by Rootzén and Zholud, their results do not actually contradict the findings presented in our original paper: whether unbounded or not, human lifespan is unlikely to greatly exceed 120 years, and the improbability of longer survival---whether it is exactly zero or merely astronomically small---acts as a de facto limit. In order to eliminate the confusion surrounding the issue, we propose the adoption of the term "limit" to denote the age at which the chance of survival is exactly zero and the term "effective limit" to denote the age at which the change of survival falls below a given threshold. Once this distinction is made, it can be demonstrated that the final result of Rootzén and Zholud is essentially a recapitulation of the main conclusion of our paper. Ultimately, much of the controversy surrounding the issue of a limit to human lifespan can be avoided by carefully reading the literature and applying statistics to practical human scales.

stat.AP↗

arXiv2018

Novel high-pressure calcium carbonates

Calcium and magnesium carbonates are believed to be the host compounds for most of the oxidized carbon in the Earth's mantle. Here, using evolutionary crystal structure prediction method USPEX, we systematically explore the MgO-CO2 and CaO-CO2 systems at pressures ranging from 0 to 160 GPa to search for thermodynamically stable magnesium and calcium carbonates. While MgCO3 is the only stable magnesium carbonate, three calcium carbonates are stable under pressure: well-known CaCO3, and newly predicted Ca3CO5 and CaC2O5. Ca3CO5 polymorphs are found to contain isolated orthocarbonate (CO4)4- tetrahedra, and are stable at relatively low pressures (>11 GPa), whereas CaC2O5 is stable above 33 GPa and its polymorphs feature polymeric motifs made of CO4 tetrahedra. Detailed analysis of chemical stability of CaCO3, Ca3CO5 and CaC2O5 in the environment typical of the Earth's lower mantle reveals that none of these compounds can exist in the Earth's lower mantle. We conclude that MgCO3 is the main host of oxidized carbon throughout the lower mantle.

cond-mat.mtrl-sci↗

arXiv2018

Spacetime as the optimal generative network of quantum states: a roadmap to QM=GR?

The idea that spacetime geometry is built from quantum entanglement has been widely accepted in the last years. But how exactly the geometry is related with quantum states is still unclear. In this note, based on the idea of deep learning, we propose a mechanism for Susskind's QM=GR hypothesis, spacetime geometry as the optimal generative network of quantum states. We speculate that the space geometry stems as a geodesic tensor network which defines the quantum state complexity of a fundamental quantum state under a given metric. Spacetime corresponds to an evolving tensor network that generates an evolutional fundamental quantum system. This mechanism provides (a) a constructive correspondence between quantum states and spacetime geometry; (b)a spacetime structure emerging from a highly constrained geodesic so that the QEC-like structure shown in AdS/CFT can be naturally realized and (c) a mechanism to derive the gravity equation from the concept of quantum state complexity. With this mechanism, spacetime can have a quantum mechanical description. We hope this may lead to another view direction to understand the basic rules of our world.

quant-ph↗

arXiv2018

Condensed matter realization of fermion quasiparticles in Minkowski space

"What is the difference between space and time?" is an ancient question that remains a matter of intense debate. In Newtonian mechanics time is absolute, while in Einstein's theory of relativity time and space combine into Minkowski spacetime. Here, we firstly propose Minkowski fermions in 2+1 dimensional Minkowski spacetime which have two space-like and one time-like momentum axes. These quasiparticles can be further classified as Klein-Gordon fermions and Dirac-Minkowski fermions according to the linearly and quadratically dispersing excitations. Realization of Dirac-Minkowski quasiparticles requires systems with particular topological nodal-line band degeneracies, such as hyperbolic nodal lines or coplanar band crossings. With the help of first-principles calculations we find that novel massless Dirac-Minkowski fermions are realized in a metastable bulk boron allotrope, Pnnm-B16.

cond-mat.mtrl-sci↗