arXiv subjects

Zhiqiang Zhang

Publications and source records attributed to Zhiqiang Zhang.

171 records · Page 10Linked to original sources

arXiv2019

Suppressing inhomogeneous broadening in a lutetium multi-ion optical clock

We demonstrate precision measurement and control of inhomogeneous broadening in a multi-ion clock consisting of three $^{176}$Lu$^+$ ions. Microwave spectroscopy between hyperfine states in the $^3D_1$ level is used to characterise differential systematic shifts between ions, most notably those associated with the electric quadrupole moment. By appropriate alignment of the magnetic field, we demonstrate suppression of these effects to the $\sim 10^{-17}$ level relative to the $^1S_0\leftrightarrow{}^3D_1$ optical transition frequency. Correlation spectroscopy on the optical transition demonstrates the feasibility of a 10s Ramsey interrogation in the three ion configuration with a corresponding projection noise limited stability of $σ(τ)=8.2\times 10^{-17}/\sqrtτ$

physics.atom-ph↗

arXiv2019

Improve Diverse Text Generation by Self Labeling Conditional Variational Auto Encoder

Diversity plays a vital role in many text generating applications. In recent years, Conditional Variational Auto Encoders (CVAE) have shown promising performances for this task. However, they often encounter the so called KL-Vanishing problem. Previous works mitigated such problem by heuristic methods such as strengthening the encoder or weakening the decoder while optimizing the CVAE objective function. Nevertheless, the optimizing direction of these methods are implicit and it is hard to find an appropriate degree to which these methods should be applied. In this paper, we propose an explicit optimizing objective to complement the CVAE to directly pull away from KL-vanishing. In fact, this objective term guides the encoder towards the "best encoder" of the decoder to enhance the expressiveness. A labeling network is introduced to estimate the "best encoder". It provides a continuous label in the latent space of CVAE to help build a close connection between latent variables and targets. The whole proposed method is named Self Labeling CVAE~(SLCVAE). To accelerate the research of diverse text generation, we also propose a large native one-to-many dataset. Extensive experiments are conducted on two tasks, which show that our method largely improves the generating diversity while achieving comparable accuracy compared with state-of-art algorithms.

stat.ML↗

arXiv2018

Semantic Human Matting

Human matting, high quality extraction of humans from natural images, is crucial for a wide variety of applications. Since the matting problem is severely under-constrained, most previous methods require user interactions to take user designated trimaps or scribbles as constraints. This user-in-the-loop nature makes them difficult to be applied to large scale data or time-sensitive scenarios. In this paper, instead of using explicit user input constraints, we employ implicit semantic constraints learned from data and propose an automatic human matting algorithm (SHM). SHM is the first algorithm that learns to jointly fit both semantic information and high quality details with deep networks. In practice, simultaneously learning both coarse semantics and fine details is challenging. We propose a novel fusion strategy which naturally gives a probabilistic estimation of the alpha matte. We also construct a very large dataset with high quality annotations consisting of 35,513 unique foregrounds to facilitate the learning and evaluation of human matting. Extensive experiments on this dataset and plenty of real images show that SHM achieves comparable results with state-of-the-art interactive matting methods.

cs.CV↗

arXiv2018

Image Matters: Visually modeling user behaviors using Advanced Model Server

In Taobao, the largest e-commerce platform in China, billions of items are provided and typically displayed with their images. For better user experience and business effectiveness, Click Through Rate (CTR) prediction in online advertising system exploits abundant user historical behaviors to identify whether a user is interested in a candidate ad. Enhancing behavior representations with user behavior images will help understand user's visual preference and improve the accuracy of CTR prediction greatly. So we propose to model user preference jointly with user behavior ID features and behavior images. However, training with user behavior images brings tens to hundreds of images in one sample, giving rise to a great challenge in both communication and computation. To handle these challenges, we propose a novel and efficient distributed machine learning paradigm called Advanced Model Server (AMS). With the well known Parameter Server (PS) framework, each server node handles a separate part of parameters and updates them independently. AMS goes beyond this and is designed to be capable of learning a unified image descriptor model shared by all server nodes which embeds large images into low dimensional high level features before transmitting images to worker nodes. AMS thus dramatically reduces the communication load and enables the arduous joint training process. Based on AMS, the methods of effectively combining the images and ID features are carefully studied, and then we propose a Deep Image CTR Model. Our approach is shown to achieve significant improvements in both online and offline evaluations, and has been deployed in Taobao display advertising system serving the main traffic.

cs.CV↗

arXiv2018

Cascaded Pyramid Network for Multi-Person Pose Estimation

The topic of multi-person pose estimation has been largely improved recently, especially with the development of convolutional neural network. However, there still exist a lot of challenging cases, such as occluded keypoints, invisible keypoints and complex background, which cannot be well addressed. In this paper, we present a novel network structure called Cascaded Pyramid Network (CPN) which targets to relieve the problem from these "hard" keypoints. More specifically, our algorithm includes two stages: GlobalNet and RefineNet. GlobalNet is a feature pyramid network which can successfully localize the "simple" keypoints like eyes and hands but may fail to precisely recognize the occluded or invisible keypoints. Our RefineNet tries explicitly handling the "hard" keypoints by integrating all levels of feature representations from the GlobalNet together with an online hard keypoint mining loss. In general, to address the multi-person pose estimation problem, a top-down pipeline is adopted to first generate a set of human bounding boxes based on a detector, followed by our CPN for keypoint localization in each human bounding box. Based on the proposed algorithm, we achieve state-of-art results on the COCO keypoint benchmark, with average precision at 73.0 on the COCO test-dev dataset and 72.1 on the COCO test-challenge dataset, which is a 19% relative improvement compared with 60.5 from the COCO 2016 keypoint challenge.Code (https://github.com/chenyilun95/tf-cpn.git) and the detection results are publicly available for further research.

cs.CV↗

arXiv2012

Identifying Brain Image Level Endophenotypes in Epilepsy

A brain wide association study (BWAS) based on the logistic regression was first developed and applied to a large population of epilepsy patients (168) and healthy controls (136). It was found that the most significant links associated with epilepsy are those bilateral links with regions mainly belonging to the default mode network and subcortex, such as amygdala, fusiform gyrus, inferior temporal gyrus, hippocampus, temporal pole, parahippocampal gyrus, insula, middle occipital gyrus, cuneus. These links were found to have much higher odd ratios than other links, and all of them showed reduced functional couplings in patients compared with controls. Interestingly, with the increasing of the seizure onset frequency or duration of illness, the functional connection between these bilateral regions became further reduced. On the other hand, as a functional compensation and brain plasticity, connections of these bilateral regions to other brain regions were abnormally enhanced and became even much stronger with the increase of the seizure onset frequency. Furthermore, patients had higher network efficiencies than healthy controls, and the seizure onset frequency was found to be positively correlated with the network efficiency. A negative correlation between the bilateral connection and the network efficiency was also observed. To further validate our findings, we then employed our BWAS results in discriminating patients from healthy controls and the leave-one-out accuracy was around 78%. Given the fact that a genome-wide association study with a large cohort has failed to identify any significant association between genes and epilepsy, our study could provide us with a set of endophenotypes for further study.

physics.med-ph↗

arXiv2010

Evolution of the Surface Structures on SrTiO$_3$(110) Tuned by Ti or Sr Concentration

The surface structure of the SrTiO$_3$(110) polar surface is studied by scanning tunneling microscopy and X-ray photoelectron spectroscopy. Monophased reconstructions in (5$\times$1), (4$\times$1), (2$\times$8), and (6$\times$8) are obtained, respectively, and the evolution between these phases can be tuned reversibly by adjusting the Ar$^{+}$ sputtering dose or the amount of Sr/Ti evaporation. Upon annealing, the surface reaches the thermodynamic equilibrium that is determined by the surface metal concentration. The different electronic structures and absorption behaviors of the surface with different reconstructions are investigated.

cond-mat.mtrl-sci↗

arXiv2010

On the parity complexity measures of Boolean functions

The parity decision tree model extends the decision tree model by allowing the computation of a parity function in one step. We prove that the deterministic parity decision tree complexity of any Boolean function is polynomially related to the non-deterministic complexity of the function or its complement. We also show that they are polynomially related to an analogue of the block sensitivity. We further study parity decision trees in their relations with an intermediate variant of the decision trees, as well as with communication complexity.

cs.CC↗

arXiv2008

Communication Complexities of XOR functions

We call $F:\{0, 1\}^n\times \{0, 1\}^n\to\{0, 1\}$ a symmetric XOR function if for a function $S:\{0, 1, ..., n\}\to\{0, 1\}$, $F(x, y)=S(|x\oplus y|)$, for any $x, y\in\{0, 1\}^n$, where $|x\oplus y|$ is the Hamming weight of the bit-wise XOR of $x$ and $y$. We show that for any such function, (a) the deterministic communication complexity is always $Θ(n)$ except for four simple functions that have a constant complexity, and (b) up to a polylog factor, the error-bounded randomized and quantum communication complexities are $Θ(r_0+r_1)$, where $r_0$ and $r_1$ are the minimum integers such that $r_0, r_1\leq n/2$ and $S(k)=S(k+2)$ for all $k\in[r_0, n-r_1)$.

quant-ph↗