arXiv subjects

Jun Li

Publications and source records attributed to Jun Li.

At least 757 records · Page 42Linked to original sources

arXiv2019

Weighted graphlets and deep neural networks for protein structure classification

As proteins with similar structures often have similar functions, analysis of protein structures can help predict protein functions and is thus important. We consider the problem of protein structure classification, which computationally classifies the structures of proteins into pre-defined groups. We develop a weighted network that depicts the protein structures, and more importantly, we propose the first graphlet-based measure that applies to weighted networks. Further, we develop a deep neural network (DNN) composed of both convolutional and recurrent layers to use this measure for classification. Put together, our approach shows dramatic improvements in performance over existing graphlet-based approaches on 36 real datasets. Even comparing with the state-of-the-art approach, it almost halves the classification error. In addition to protein structure networks, our weighted-graphlet measure and DNN classifier can potentially be applied to classification of other weighted networks in computational biology as well as in other domains.

stat.ML↗

arXiv2019

Subsystem-based Approach to Scalable Quantum Optimal Control

The development of quantum control methods is an essential task for emerging quantum technologies. In general, the process of optimizing quantum controls scales very unfavorably in system size due to the exponential growth of the Hilbert space dimension. Here, I present a scalable subsystem-based method for quantum optimal control on a large quantum system. The basic idea is to cast the original problem as a robust control problem on subsystems with requirement of robustness in respect of inter-subsystem couplings, thus enabling a drastic reduction of problem size. The method is in particular suitable for target quantum operations that are local, e.g., elementary quantum gates. As an illustrative example, I employ it to tackle the demanding task of pulse searching on a 12-spin coupled system, achieving substantial reduction of memory and time costs. This work has significant implications for coherent quantum engineering on near-term intermediate-scale quantum devices.

quant-ph↗

arXiv2019

Weakly Supervised Energy-Based Learning for Action Segmentation

This paper is about labeling video frames with action classes under weak supervision in training, where we have access to a temporal ordering of actions, but their start and end frames in training videos are unknown. Following prior work, we use an HMM grounded on a Gated Recurrent Unit (GRU) for frame labeling. Our key contribution is a new constrained discriminative forward loss (CDFL) that we use for training the HMM and GRU under weak supervision. While prior work typically estimates the loss on a single, inferred video segmentation, our CDFL discriminates between the energy of all valid and invalid frame labelings of a training video. A valid frame labeling satisfies the ground-truth temporal ordering of actions, whereas an invalid one violates the ground truth. We specify an efficient recursive algorithm for computing the CDFL in terms of the logadd function of the segmentation energy. Our evaluation on action segmentation and alignment gives superior results to those of the state of the art on the benchmark Breakfast Action, Hollywood Extended, and 50Salads datasets.

cs.CV↗

arXiv2019

Variable selection with false discovery rate control in deep neural networks

Deep neural networks (DNNs) are famous for their high prediction accuracy, but they are also known for their black-box nature and poor interpretability. We consider the problem of variable selection, that is, selecting the input variables that have significant predictive power on the output, in DNNs. We propose a backward elimination procedure called SurvNet, which is based on a new measure of variable importance that applies to a wide variety of networks. More importantly, SurvNet is able to estimate and control the false discovery rate of selected variables, while no existing methods provide such a quality control. Further, SurvNet adaptively determines how many variables to eliminate at each step in order to maximize the selection efficiency. To study its validity, SurvNet is applied to image data and gene expression data, as well as various simulation datasets.

stat.ML↗

arXiv2019

Fine Tuning Problem of the Cosmological Constant in a Generalized Randall-Sundrum Model

To solve the cosmological constant fine tuning problem, we investigate a $(n+1)$-dimensional generalized Randall-Sundrum brane world scenario with two $(n-1)$-branes instead of two 3-branes. Adopting an anisotropic metric ansatz, we obtain the positive effective cosmological constant $Ω_{eff}$ of order $10^{-124}$ and only require a solution $\simeq50-80$. Meanwhile, both the visible and hidden branes are stable because their tensions are positive. Therefore, the fine tuning problem can be solved quite well. Furthermore, the Hubble parameter $H_{1}(z)$ as a function of redshift $z$ is in good agreement with cosmic chronometers dataset. The evolution of the universe naturally shifts from deceleration to acceleration. This demonstrates that the evolution of the universe is intrinsically an extra-dimensional phenomenon. It can be seen as a dynamic model of dark energy which is driven by the evolution of the extra dimensions on the brane.

gr-qc↗

arXiv2019

Cross-X Learning for Fine-Grained Visual Categorization

Recognizing objects from subcategories with very subtle differences remains a challenging task due to the large intra-class and small inter-class variation. Recent work tackles this problem in a weakly-supervised manner: object parts are first detected and the corresponding part-specific features are extracted for fine-grained classification. However, these methods typically treat the part-specific features of each image in isolation while neglecting their relationships between different images. In this paper, we propose Cross-X learning, a simple yet effective approach that exploits the relationships between different images and between different network layers for robust multi-scale feature learning. Our approach involves two novel components: (i) a cross-category cross-semantic regularizer that guides the extracted features to represent semantic parts and, (ii) a cross-layer regularizer that improves the robustness of multi-scale features by matching the prediction distribution across multiple layers. Our approach can be easily trained end-to-end and is scalable to large datasets like NABirds. We empirically analyze the contributions of different components of our approach and demonstrate its robustness, effectiveness and state-of-the-art performance on five benchmark datasets. Code is available at \url{https://github.com/cswluo/CrossX}.

cs.CV↗

arXiv2019

Clustering with Outlier Removal

Cluster analysis and outlier detection are strongly coupled tasks in data mining area. Cluster structure can be easily destroyed by few outliers; on the contrary, outliers are defined by the concept of cluster, which are recognized as the points belonging to none of the clusters. Unfortunately, most existing studies do not notice the coupled relationship between these two task and handle them separately. In light of this, we consider the joint cluster analysis and outlier detection problem, and propose the Clustering with Outlier Removal (COR) algorithm. Generally speaking, the original space is transformed into the binary space via generating basic partitions in order to define clusters. Then an objective function based Holoentropy is designed to enhance the compactness of each cluster with a few outliers removed. With further analyses on the objective function, only partial of the problem can be handled by K-means optimization. To provide an integrated solution, an auxiliary binary matrix is nontrivally introduced so that COR completely and efficiently solves the challenging problem via a unified K-means-- with theoretical supports. Extensive experimental results on numerous data sets in various domains demonstrate the effectiveness and efficiency of COR significantly over state-of-the-art methods in terms of cluster validity and outlier detection. Some key factors in COR are further analyzed for practical use. Finally, an application on flight trajectory is provided to demonstrate the effectiveness of COR in the real-world scenario.

cs.LG↗

arXiv2019

Anisotropic magnetocaloric effect in Fe$_{3-x}$GeTe$_2$

We present a comprehensive study on anisotropic magnetocaloric porperties of the van der Waals weak-itinerant ferromagnet Fe$_{3-x}$GeTe$_2$ that features gate-tunable room-temperature ferromagnetism in few-layer device. Intrinsic magnetocrystalline anisotropy is observed to be temperature-dependent and most likely favors the long-range magnetic order in thin Fe$_{3-x}$GeTe$_2$ crsytal. The magnetic entropy change $-ΔS_M$ also reveals an anisotropic characteristic between $H // ab$ and $H // c$, which could be well scaled into a universal curve. The peak value $-ΔS_M^{max}$ of 1.20 J kg$^{-1}$ K$^{-1}$ and the corresponding adiabatic temperature change $ΔT_{ad}$ of 0.66 K are deduced from heat capacity with out-of-plane field change of 5 T. By fitting of the field-dependent parameters of $-ΔS_M^{max}$ and the relative cooling power RCP, it gives $-ΔS_M^{max} \propto H^n$ with $n = 0.603(6)$ and $RCP \propto H^m$ with $m = 1.20(1)$ when $H // c$. Given the high and tunable $T_c$, Fe$_{3-x}$GeTe$_2$ crystals are of interest for fabricating the heterostructure-based spintronics device.

cond-mat.str-el↗

arXiv2019

Thickness-dependent magnetic order in CrI$_3$ single crystals

Two-dimensional (2D) materials with intrinsic ferromagnetism provide unique opportunity to engineer new functionalities in nano-spintronics. One such material is CrI$_3$, showing long-range magnetic order in monolayer with the Curie temperature ($T_c$) of 45 K. Here we study detailed evolution of magnetic transition and magnetic critical properties in response to systematic reduction in crystal thickness down to 50 nm. Bulk $T_c$ of 61 K is gradually suppressed to 57 K, however, the satellite transition at $T^*$ = 45 K is observed layer-independent at fixed magnetic field of 1 kOe. The origin of $T^*$ is proposed to be a crossover from pinning to depinning of magnetic domain walls. The reduction of thickness facilitates a field-driven metamagnetic transition around 20 kOe with out-of-plane field, in contrast to the continuous changes with in-plane field. The critical analysis around $T_c$ elucidates the mean-field type interactions in microscale-thick CrI$_3$.

cond-mat.str-el↗

arXiv2019

The theory of N-Mixed-Spin-P fields

This is the first part of the project toward proving the BCOV's Feymann graph sum formula of all genera Gromov-Witten invariants of quintic Calabi-Yau threefolds. In this paper, we introduce the notion of N-Mixed-Spin-P fields, construct their moduli spaces, their virtual cycles, their virtual localization formulas, and a vanishing result associated with irregular graphs.

math.AG↗

arXiv2019

Formation of Graphene atop a Si adlayer on the C-face of SiC

The structure of the SiC(000-1) surface, the C-face of the {0001} SiC surfaces, is studied as a function of temperature and of pressure in a gaseous environment of disilane (Si2H6). Various surface reconstructions are observed, both with and without the presence of an overlying graphene layer (which spontaneously forms at sufficiently high temperatures). Based on cross-sectional scanning transmission electron microscopy measurements, the interface structure that forms in the presence of the graphene is found to contain 1.4 - 1.7 monolayers (ML) of Si, a somewhat counter-intuitive result since, when the graphene forms, the system is actually under C-rich conditions. Using ab initio thermodynamics, it is demonstrated that there exists a class of Si-rich surfaces containing about 1.3 ML of Si that are stable on the surface (even under C-rich conditions) at temperatures above about 400 K. The structures that thus form consist of Si adatoms atop a Si adlayer on the C-face of SiC, with or without the presence of overlying graphene.

cond-mat.mes-hall↗

arXiv2019

Cost-Driven Offloading for DNN-based Applications over Cloud, Edge and End Devices

Currently, deep neural networks (DNNs) have achieved a great success in various applications. Traditional deployment for DNNs in the cloud may incur a prohibitively serious delay in transferring input data from the end devices to the cloud. To address this problem, the hybrid computing environments, consisting of the cloud, edge and end devices, are adopted to offload DNN layers by combining the larger layers (more amount of data) in the cloud and the smaller layers (less amount of data) at the edge and end devices. A key issue in hybrid computing environments is how to minimize the system cost while accomplishing the offloaded layers with their deadline constraints. In this paper, a self-adaptive discrete particle swarm optimization (PSO) algorithm using the genetic algorithm (GA) operators was proposed to reduce the system cost caused by data transmission and layer execution. This approach considers the characteristics of DNNs partitioning and layers offloading over the cloud, edge and end devices. The mutation operator and crossover operator of GA were adopted to avert the premature convergence of PSO, which distinctly reduces the system cost through enhanced population diversity of PSO. The proposed offloading strategy is compared with benchmark solutions, and the results show that our strategy can effectively reduce the cost of offloading for DNN-based applications over the cloud, edge and end devices relative to the benchmarks.

cs.DC↗

arXiv2019

Measuring the tilt of primordial gravitational-wave power spectrum from observations

Primordial gravitational waves generated during inflation lead to the B-mode polarization in the cosmic microwave background and a stochastic gravitational wave background in the Universe. We will explore the current constraint on the tilt of primordial gravitational-wave spectrum, and forecast how the future observations can improve the current constraint.

astro-ph.CO↗

arXiv2019

Generative Steganography by Sampling

In this paper, a novel data-driven information hiding scheme called generative steganography by sampling (GSS) is proposed. Unlike in traditional modification-based steganography, in our method the stego image is directly sampled by a powerful generator: no explicit cover is used. Both parties share a secret key used for message embedding and extraction. The Jensen-Shannon divergence is introduced as a new criterion for evaluating the security of generative steganography. Based on these principles, we propose a simple practical generative steganography method that uses semantic image inpainting. The message is written in advance to an uncorrupted region that needs to be retained in the corrupted image. Then, the corrupted image with the secret message is fed into a Generator trained by a generative adversarial network (GAN) for semantic completion. Message loss and prior loss terms are proposed for penalizing message extraction error and unrealistic stego image. In our design, we first train a generator whose training target is the generation of new data samples from the same distribution as that of existing training data. Next, for the trained generator, backpropagation to the message and prior loss are introduced to optimize the coding of the input noise data for the generator. The presented experiments demonstrate the potential of the proposed framework based on both qualitative and quantitative evaluations of the generated stego images.

cs.MM↗

arXiv2019

Preliminary study on the modal decomposition of Hermite Gaussian beams via deep learning

The Hermite-Gaussian (HG) modes make up a complete and orthonormal basis, which have been extensively used to describe optical fields. Here, we demonstrate, for the first time to our knowledge, deep learning-based modal decomposition (MD) of HG beams. This method offers a fast, economical and robust way to acquire both the power content and phase information through a single-shot beam intensity image, which will be beneficial for the beam shaping, beam quality assessment, studies of resonator perturbations, and other further research on the HG beams.

physics.optics↗

arXiv2019

On the mathematics and physics of Mixed Spin P-Fields

We outline various developments of affine and general Landau Ginzburg models in physics. We then describe the A-twisting and coupling to gravity in terms of Algebraic Geometry. We describe constructions of various path integral measures (virtual fundamental class) using the algebro-geometric technique of cosection localization, culminating in the theory of ``Mixed Spin P (MSP) fields" developed by the authors.

math.AG↗

arXiv2019

Deep learning enabled superfast and accurate M^2 evaluation for fiber beams

We introduce deep learning technique to predict the beam propagation factor M^2 of the laser beams emitting from few-mode fiber for the first time, to the best of our knowledge. The deep convolutional neural network (CNN) is trained with paired data of simulated near-field beam patterns and their calculated M^2 value, aiming at learning a fast and accurate mapping from the former to the latter. The trained deep CNN can then be utilized to evaluate M^2 of the fiber beams from single beam patterns. The results of simulated testing samples have shown that our scheme can achieve an averaged prediction error smaller than 2% even when up to 10 eigenmodes are involved in the fiber. The error becomes slightly larger when heavy noises are added into the input beam patterns but still smaller than 2.5%, which further proves the accuracy and robustness of our method. Furthermore, the M^2 estimation takes only about 5 ms for a prepared beam pattern with one forward pass, which can be adopted for real-time M^2 determination with only one supporting Charge-Coupled Device (CCD). The experimental results further prove the feasibility of our scheme. Moreover, the method we proposed can be confidently extended to other kinds of beams provided that adequate training samples are accessible. Deep learning paves the way to superfast and accurate M^2 evaluation with very low experimental efforts.

eess.IV↗

arXiv2019

Network analysis of synonymous codon usage

Most amino acids are encoded by multiple synonymous codons. For an amino acid, some of its synonymous codons are used much more rarely than others. Analyses of positions of such rare codons in protein sequences revealed that rare codons can impact co-translational protein folding and that positions of some rare codons are evolutionary conserved. Analyses of positions of rare codons in proteins' 3-dimensional structures, which are richer in biochemical information than sequences alone, might further explain the role of rare codons in protein folding. We analyze a protein set recently annotated with codon usage information, considering non-redundant proteins with sufficient structural information. We model the proteins' structures as networks and study potential differences between network positions of amino acids encoded by evolutionary conserved rare, evolutionary non-conserved rare, and commonly used codons. In 84% of the proteins, at least one of the three codon categories occupies significantly more or less network-central positions than the other codon categories. Different protein groups showing different codon centrality trends (i.e., different types of relationships between network positions of the three codon categories) are enriched in different biological functions, implying the existence of a link between codon usage, protein folding, and protein function.

q-bio.MN↗