arXiv subjects

Xiaofei Zhang

Publications and source records attributed to Xiaofei Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2023

Efficient Quality-Diversity Optimization through Diverse Quality Species

A prevalent limitation of optimizing over a single objective is that it can be misguided, becoming trapped in local optimum. This can be rectified by Quality-Diversity (QD) algorithms, where a population of high-quality and diverse solutions to a problem is preferred. Most conventional QD approaches, for example, MAP-Elites, explicitly manage a behavioral archive where solutions are broken down into predefined niches. In this work, we show that a diverse population of solutions can be found without the limitation of needing an archive or defining the range of behaviors in advance. Instead, we break down solutions into independently evolving species and use unsupervised skill discovery to learn diverse, high-performing solutions. We show that this can be done through gradient-based mutations that take on an information theoretic perspective of jointly maximizing mutual information and performance. We propose Diverse Quality Species (DQS) as an alternative to archive-based QD algorithms. We evaluate it over several simulated robotic environments and show that it can learn a diverse set of solutions from varying species. Furthermore, our results show that DQS is more sample-efficient and performant when compared to other QD algorithms. Relevant code and hyper-parameters are available at: https://github.com/rwickman/NEAT_RL.

cs.LG↗

arXiv2023

UniFed: A Unified Framework for Federated Learning on Non-IID Image Features

How to tackle non-iid data is a crucial topic in federated learning. This challenging problem not only affects training process, but also harms performance of clients not participating in training. Existing literature mainly focuses on either side, yet still lacks a unified solution to handle these two types (internal and external) of clients in a joint way. In this work, we propose a unified framework to tackle the non-iid issues for internal and external clients together. Firstly, we propose to use client-specific batch normalization in either internal or external clients to alleviate feature distribution shifts incurred by non-iid data. Then we present theoretical analysis to demonstrate the benefits of client-specific batch normalization. Specifically, we show that our approach promotes convergence speed for federated training and yields lower generalization error bound for external clients. Furthermore, we use causal reasoning to form a causal view to explain the advantages of our framework. At last, we conduct extensive experiments on natural and medical images to evaluate our method, where our method achieves state-of-the-art performance, faster convergence, and shows good compatibility. We also performed comprehensive analytical studies on a real-world medical dataset to demonstrate the effectiveness.

cs.LG↗

arXiv2023

Learning-Based Data Storage [Vision] (Technical Report)

Deep neural network (DNN) and its variants have been extensively used for a wide spectrum of real applications such as image classification, face/speech recognition, fraud detection, and so on. In addition to many important machine learning tasks, as artificial networks emulating the way brain cells function, DNNs also show the capability of storing non-linear relationships between input and output data, which exhibits the potential of storing data via DNNs. We envision a new paradigm of data storage, "DNN-as-a-Database", where data are encoded in well-trained machine learning models. Compared with conventional data storage that directly records data in raw formats, learning-based structures (e.g., DNN) can implicitly encode data pairs of inputs and outputs and compute/materialize actual output data of different resolutions only if input data are provided. This new paradigm can greatly enhance the data security by allowing flexible data privacy settings on different levels, achieve low space consumption and fast computation with the acceleration of new hardware (e.g., Diffractive Neural Network and AI chips), and can be generalized to distributed DNN-based storage/computing. In this paper, we propose this novel concept of learning-based data storage, which utilizes a learning structure called learning-based memory unit (LMU), to store, organize, and retrieve data. As a case study, we use DNNs as the engine in the LMU, and study the data capacity and accuracy of the DNN-based data storage. Our preliminary experimental results show the feasibility of the learning-based data storage by achieving high (100%) accuracy of the DNN storage. We explore and design effective solutions to utilize the DNN-based data storage to manage and query relational tables. We discuss how to generalize our solutions to other data types (e.g., graphs) and environments such as distributed DNN storage/computing.

cs.DB↗

arXiv2023

A Generic Graph Sparsification Framework using Deep Reinforcement Learning

The interconnectedness and interdependence of modern graphs are growing ever more complex, causing enormous resources for processing, storage, communication, and decision-making of these graphs. In this work, we focus on the task graph sparsification: an edge-reduced graph of a similar structure to the original graph is produced while various user-defined graph metrics are largely preserved. Existing graph sparsification methods are mostly sampling-based, which introduce high computation complexity in general and lack of flexibility for a different reduction objective. We present SparRL, the first generic and effective graph sparsification framework enabled by deep reinforcement learning. SparRL can easily adapt to different reduction goals and promise graph-size-independent complexity. Extensive experiments show that SparRL outperforms all prevailing sparsification methods in producing high-quality sparsified graphs concerning a variety of objectives.

cs.LG↗

arXiv2022

Channel Modeling for UAV-to-Ground Communications with Posture Variation and Fuselage Scattering Effect

Unmanned aerial vehicle (UAV)-to-ground (U2G) channel models play a pivotal role for reliable communications between UAV and ground terminal. This paper proposes a three-dimensional (3D) non-stationary hybrid model including both large-scale and small-scale fading for U2G multiple-input-multiple-output (MIMO) channels. Distinctive channel characteristics under U2G scenarios, i.e., 3D trajectory and posture of UAV, fuselage scattering effect (FSE), and posture variation fading (PVF), are incorporated into the proposed model. The channel parameters, i.e., path loss (PL), shadow fading (SF), path delay, and path angle, are generated incorporating machine learning (ML) and ray tracing (RT) techniques to capture the structure-related characteristics. In order to guarantee the physical continuity of channel parameters such as Doppler phase and path power, the time evolution methods of inter- and intra- stationary intervals are proposed. Key statistical properties , i.e., temporal autocorrection function (ACF), power delay profile (PDP), level crossing rate (LCR), average fading duration (AFD), and stationary interval (SI) are given, and the impact of the change of fuselage and posture variation is analyzed. It is demonstrated that both posture variation and fuselage scattering have crucial effects on channel characteristics. The validity and practicability of the proposed model are verified by comparing the simulation results with the measured ones.

eess.SP↗

arXiv2022

Edge Semantic Cognitive Intelligence for 6G Networks: Novel Theoretical Models, Enabling Framework, and Typical Applications

Edge intelligence is anticipated to underlay the pathway to connected intelligence for 6G networks, but the organic confluence of edge computing and artificial intelligence still needs to be carefully treated. To this end, this article discusses the concepts of edge intelligence from the semantic cognitive perspective. Two instructive theoretical models for edge semantic cognitive intelligence (ESCI) are first established. Afterwards, the ESCI framework orchestrating deep learning with semantic communication is discussed. Two representative applications are present to shed light on the prospect of ESCI in 6G networks. Some open problems are finally listed to elicit the future research directions of ESCI.

eess.SP↗

arXiv2022

Enabling Heterogeneous Catalysis to Achieve Carbon Neutrality: Directional Catalytic Conversion of CO$_2$ into Carboxylic Acids

The increase in anthropogenic carbon dioxide (CO$_2$) emissions has exacerbated the deterioration of the global environment, which should be controlled to achieve carbon neutrality. Central to the core goal of achieving carbon neutrality is the utilization of CO$_2$ under economic and sustainable conditions. Recently, the strong need for carbon neutrality has led to a proliferation of studies on the direct conversion of CO$_2$ into carboxylic acids, which could effectively alleviate CO$_2$ emissions and create high-value chemicals. The purpose of this review is to present the application prospects of carboxylic acids and the basic principles of CO$_2$ conversion into carboxylic acids through photo-, electric-, and thermal catalysis. Special attention is focused on the regulation strategy of the activity of abundant catalysts at the molecular level, inspiring the preparation of high-performance catalysts. In addition, theoretical calculation, advanced technologies, and numerous typical examples are introduced to elaborate on the corresponding process and influencing factors of catalytic activity. Finally, challenges and prospects are provided for the future development of this field. It is hoped that this review contributes to a deeper understanding of the conversion of CO$_2$ into carboxylic acids and inspires more innovative breakthroughs

physics.chem-ph↗

arXiv2022

End-to-end translation of human neural activity to speech with a dual-dual generative adversarial network

In a recent study of auditory evoked potential (AEP) based brain-computer interface (BCI), it was shown that, with an encoder-decoder framework, it is possible to translate human neural activity to speech (T-CAS). However, current encoder-decoder-based methods achieve T-CAS often with a two-step method where the information is passed between the encoder and decoder with a shared dimension reduction vector, which may result in a loss of information. A potential approach to this problem is to design an end-to-end method by using a dual generative adversarial network (DualGAN) without dimension reduction of passing information, but it cannot realize one-to-one signal-to-signal translation (see Fig.1 (a) and (b)). In this paper, we propose an end-to-end model to translate human neural activity to speech directly, create a new electroencephalogram (EEG) datasets for participants with good attention by design a device to detect participants' attention, and introduce a dual-dual generative adversarial network (Dual-DualGAN) (see Fig. 1 (c) and (d)) to address an end-to-end translation of human neural activity to speech (ET-CAS) problem by group labelling EEG signals and speech signals, inserting a transition domain to realize cross-domain mapping. In the transition domain, the transition signals are cascaded by the corresponding EEG and speech signals in a certain proportion, which can build bridges for EEG and speech signals without corresponding features, and realize one-to-one cross-domain EEG-to-speech translation. The proposed method can translate word-length and sentence-length sequences of neural activity to speech. Experimental evaluation has been conducted to show that the proposed method significantly outperforms state-of-the-art methods on both words and sentences of auditory stimulus.

cs.SD↗

arXiv2021

One to Multiple Mapping Dual Learning: Learning Multiple Sources from One Mixed Signal

Single channel blind source separation (SCBSS) refers to separate multiple sources from a mixed signal collected by a single sensor. The existing methods for SCBSS mainly focus on separating two sources and have weak generalization performance. To address these problems, an algorithm is proposed in this paper to separate multiple sources from a mixture by designing a parallel dual generative adversarial Network (PDualGAN) that can build the relationship between a mixture and the corresponding multiple sources to realize one-to-multiple cross-domain mapping. This algorithm can be applied to any mixed model such as linear instantaneous mixed model and convolutional mixed model. Besides, one-to-multiple datasets are created which including the mixtures and corresponding sources for this study. The experiment was carried out on four different datasets and tested with signals mixed in different proportions. Experimental results show that the proposed algorithm can achieve high performance in peak signal-to-noise ratio (PSNR) and correlation, which outperforms state-of-the-art algorithms.

cs.LG↗

arXiv2021

FedBN: Federated Learning on Non-IID Features via Local Batch Normalization

The emerging paradigm of federated learning (FL) strives to enable collaborative training of deep models on the network edge without centrally aggregating raw data and hence improving data privacy. In most cases, the assumption of independent and identically distributed samples across local clients does not hold for federated learning setups. Under this setting, neural network training performance may vary significantly according to the data distribution and even hurt training convergence. Most of the previous work has focused on a difference in the distribution of labels or client shifts. Unlike those settings, we address an important problem of FL, e.g., different scanners/sensors in medical imaging, different scenery distribution in autonomous driving (highway vs. city), where local clients store examples with different distributions compared to other clients, which we denote as feature shift non-iid. In this work, we propose an effective method that uses local batch normalization to alleviate the feature shift before averaging models. The resulting scheme, called FedBN, outperforms both classical FedAvg, as well as the state-of-the-art for non-iid data (FedProx) on our extensive experiments. These empirical results are supported by a convergence analysis that shows in a simplified setting that FedBN has a faster convergence rate than FedAvg. Code is available at https://github.com/med-air/FedBN.

cs.LG↗

arXiv2019

Systematic external evaluation of published population pharmacokinetic models for tacrolimus in adult liver transplant recipients

Background:Diverse tacrolimus population pharmacokinetic models in adult liver transplant recipients have been established to describe the PK characteristics of tacrolimus in the last two decades. However, their extrapolated predictive performance remains unclear.Therefore,in this study,we aimed to evaluate their external predictability and identify their potential influencing factors. Methods:The external predictability of each selected popPK model was evaluated using an independent dataset of 84 patients with 572 trough concentrations prospectively collected from Huashan Hospital. Prediction and simulation based diagnostics and Bayesian forecasting were conducted to evaluate model predictability. Furthermore, the effect of model structure on the predictive performance was investigated.Results:Sixteen published popPK models were assessed. In prediction-based diagnostics,the prediction error within 30% was below 50% in all the published models. The simulation based normalised prediction distribution error test and visual predictive check indicated large discrepancies between the observations and simulations in most of the models. Bayesian forecasting showed improvement in model predictability with two to three prior observations. Additionally, the predictive performance of the nonlinear Michaelis Menten model was superior to that of linear compartment models,indicating the underlying nonlinear kinetics of tacrolimus in liver transplant recipients.Conclusions:The published models performed inadequately in prediction and simulation based diagnostics. Bayesian forecasting may improve the predictive performance of the models. Furthermore, nonlinear kinetics of tacrolimus may be mainly caused by the properties of the drug itself, and incorporating nonlinear kinetics may be considered to improve model predictability.

q-bio.QM↗

arXiv2018

Multi-cavity ultra-stable laser towards 10-18

In this letter, we demonstrate a technique of making an ultrastable laser referenced to a multi-cavity, corresponding to a lower thermal noise limit due to the larger equivalent beam size. The multi-cavity consists of several pairs of mirrors and a common spacer. We can stabilize the laser frequencies on these cavities, and average the laser frequencies with synthesizing technique. In comparison with a single cavity system, relative frequency instability of the synthesized laser can be improved by a factor of the squire root of the cavity number (n). In addition, we perform an experiment to simulate a two-cavity system. Experimental results show that frequency instability of the synthesized laser is improved by a factor of 1.4, and discrimination of the laser frequency instability, introduced by the process of lasers synthesizing, is negligible, and can reach a floor at low level 10-18 limited by noise of currently used signal generators. This technique is comparable with other techniques; thus, it can gain a factor of the squre root of n on the frequency instability of an ultrastable laser to an unprecedented level.

physics.ins-det↗

arXiv2017

Analytical results for a parity-time symmetric two-level system under synchronous combined modulations

We propose a simple method of combined synchronous modulations to generate the analytically exact solutions for a parity-time symmetric two-level system. Such exact solutions are expressible in terms of simple elementary functions and helpful for illuminating some generalizations of appealing concepts originating in the Hermitian system. Some intriguing physical phenomena, such as stabilization of a non-Hermitian system by periodic driving, non-Hermitian analogs of coherent destruction of tunneling (CDT) and complete population inversion (CPI), are demonstrated analytically and confirmed numerically. In addition, by using these exact solutions we derive a pulse area theorem for such non-Hermitian CPI in the parity-time symmetric two-level system. Our results may provide an additional possibility for pulse manipulation and coherent control of the parity-time symmetric two-level system.

quant-ph↗

arXiv2016

Subharmonic Solutions and Minimal Periodic Solutions of First-order Hamiltonian Systems with Anisotropic Growth

Using a homologically link theorem in variational theory and iteration inequalities of Maslov-type index, we show the existence of a sequence of subharmonic solutions of non-autonomous Hamiltonian systems with the Hamiltonian functions satisfying some anisotropic growth conditions, i.e., the Hamiltonian functions may have simultaneously, in different components, superquadratic, subquadratic and quadratic behaviors. Moreover, we also consider the minimal period problem of some autonomous Hamiltonian systems with anisotropic growth.

math.DS↗

arXiv2015

Optimization Design and Analysis of Systematic LT codes over AWGN Channel

In this paper, we study systematic Luby Transform (SLT) codes over additive white Gaussian noise (AWGN) channel. We introduce the encoding scheme of SLT codes and give the bipartite graph for iterative belief propagation (BP) decoding algorithm. Similar to low-density parity-check codes, Gaussian approximation (GA) is applied to yield asymptotic performance of SLT codes. Recent work about SLT codes has been focused on providing better encoding and decoding algorithms and design of degree distributions. In our work, we propose a novel linear programming method to optimize the degree distribution. Simulation results show that the proposed distributions can provide better bit-error-ratio (BER) performance. Moreover, we analyze the lower bound of SLT codes and offer closed form expressions.

cs.IT↗

arXiv2012

Efficient Multi-way Theta-Join Processing Using MapReduce

Multi-way Theta-join queries are powerful in describing complex relations and therefore widely employed in real practices. However, existing solutions from traditional distributed and parallel databases for multi-way Theta-join queries cannot be easily extended to fit a shared-nothing distributed computing paradigm, which is proven to be able to support OLAP applications over immense data volumes. In this work, we study the problem of efficient processing of multi-way Theta-join queries using MapReduce from a cost-effective perspective. Although there have been some works using the (key,value) pair-based programming model to support join operations, efficient processing of multi-way Theta-join queries has never been fully explored. The substantial challenge lies in, given a number of processing units (that can run Map or Reduce tasks), mapping a multi-way Theta-join query to a number of MapReduce jobs and having them executed in a well scheduled sequence, such that the total processing time span is minimized. Our solution mainly includes two parts: 1) cost metrics for both single MapReduce job and a number of MapReduce jobs executed in a certain order; 2) the efficient execution of a chain-typed Theta-join with only one MapReduce job. Comparing with the query evaluation strategy proposed in [23] and the widely adopted Pig Latin and Hive SQL solutions, our method achieves significant improvement of the join processing efficiency.

cs.DB↗

arXiv2007

Interpolation maps and congruence domains for wavelet sets

It is proven that if an interpolation map between two wavelet sets preserves the union of the sets, then the pair must be an interpolation pair. We also construct an example of a pair of wavelet sets for which the congruence domains of the associated interpolation map and its inverse are equal, and yet the pair is not an interpolation pair. The first result solves affirmatively a problem that the second author had posed several years ago, and the second result solves an intriguing problem of D. Han. The key to this counterexample is a special technical lemma on constructing wavelet sets. Several other applications of this result are also given. In addition, some problems are posed. We also take the opportunity to give some general exposition on wavelet sets and operator-theoretic interpolation of wavelets.

math.FA↗

arXiv2004

Probing small-$x$ gluons by low-mass Drell-Yan pairs at colliders

The transverse-momentum ($Q_T$) distribution of low-mass Drell-Yan pairs is calculated in QCD perturbation theory with all-order resummation of $α_s (α_s \ln(Q^2_T/Q^2))^n$ type terms. We demonstrate that the rapidity distribution of low-mass Drell-Yan pairs at large-enough transverse momentum is an advantageous source of constraints on the gluon distribution and its nuclear dependence. We argue that low-mass Drell-Yan pairs in the forward region provide a good and clean probe of small-$x$ gluons at collider energies.

hep-ph↗