2022年人工智能理论与算法学术研讨会

 (2022年12月3日，开封)

 

 

 

一、会议宗旨

由河南大学数学与统计学院主办，“河南省应用数学中心（河南大学）”与“河南省人工智能理论及算法工程研究中心”联合承办的“2022年人工智能理论与算法学术研讨会”将于12月3日在河南大学（金明校区）举行。会议围绕新一代人工智能的数学基础这一主题，开展关于统计学理论、优化算法、科学计算、机器学习等方面的探讨，展示最新成果、研讨热点问题、展望发展前景。会议致力于促进学者之间的交流，促进人工智能领域的发展。

二、会议主题

 包括但不限于机器学习、高维统计推断、数据挖掘、生物统计、贝叶斯分析、优化算法、密码与编码、生物大数据分析、智能信息处理、智慧医疗成像等。

三、邀请报告人

陈花萍  （河南大学）

葛志昊  （河南大学）

李凌霄  （河南大学）

庞志峰  （河南大学）

苏丽敏  （河南大学）

肖运海  （河南大学）

辛  欣  （河南大学）

薛留根  （河南大学）

杨利军  （河南大学）

杨晓慧  （河南大学）

杨翔宇  （河南大学）

郑  晨  （河南大学）

邹广安  （河南大学）

四、组织委员会

主  任：韩小森

委  员：肖运海、杨晓慧

五、会议安排与注意事项

(1)    会议时间：12月3日（周六）全天

(2)    本次会议采取线上、线下同步进行

(3)    线下会场：河南大学数学中心会议室（八大街）

(4)    采取邀请报告形式，每人报告时间30分钟

主办： 河南大学数学与统计学院

承办：河南省应用数学中心（河南大学）

河南省人工智能理论及算法工程研究中心

     

                                              

2022年11月24日

 

 

会议日程

 

Empirical Likelihood in Semiparametric Models

薛留根

In this talk, we discuss the empirical likelihood based inference problem in semiparametric models. Firstly, we investigate the empirical likelihood based inference for the parameters in a partially linear single-index model. we propose a bias correction method to achieve that the empirical likelihood ratio has standard chi-square limit. Secondly, we investigate the empirical likelihood-based inference for a varying coefficient model with longitudinal data. we propose three empirical likelihood ratios: the naive empirical likelihood ratio, the mean-corrected empirical likelihood ratio and the residual-adjusted empirical likelihood ratio, and show that these ratios have chi-square limits. In addition, when some components are of particular interest, we suggest the mean-corrected and residual-adjusted partial empirical likelihood ratios for the construction of the confidence regions/intervals. A simulation study is undertaken to compare the empirical likelihood and the normal approximation methods in terms of coverage accuracies and average areas/widths of confidence regions/intervals. An example in epidemiology is used for illustration.

 

Multiphysics Finite Element Method for Thermo-Poroelasticity and PINNs

葛志昊

In this talk, I will introduce the multiphysics finite element method for the quasi-static thermo-poroelasticity model, which can overcome two locking. And I will talk about the stability analysis of the above proposed method the optimal convergence order, and show some numerical examples to verify the theoretical results. Also, I will introduce the PINNs for poroelasticity model.

 

A Conway-Maxwell-Poisson-Binomial AR(1) Model for Bounded Time Series Data

陈花萍

Binomial autoregressive models are frequently used for modelling bounded time series counts, which is the number of the occurrence in n independent units. However, they are not well developed for more complex bounded time series counts of the occurrence of n exchangeable and dependent units, which are becoming increasingly common in practice. To fill this gap, this paper first constructs an exchangeable Conway-Maxwell-Poisson-binomial (CMPB) thinning operator and then establishes the Conway-Maxwell-Poisson-binomial AR (CMPBAR) model. Two key features of the proposed model are that it remains the autoregressive structure of the classic BAR(1) model, but also allows to analyze bounded data with under-dispersion, equi-dispersion and over-dispersion. We discuss the conditional maximum likelihood (CML) estimate of model’s parameters and establish the asymptotic normality of the CML estimator. In a simulation study, the boxplots illustrate that the CML estimator is consistence and the qqplots show that the CML estimator is asymptotic normality. In the real data example, our model takes the smaller AIC and BIC than its main competitors.

 

多物理场耦合模型的数值解法及应用

邹广安

本报告主要研究几类多物理场流体耦合模型的有限元解耦算法，讨论了全离散算法的能量稳定性和误差估计，利用数值算例验证了算法的有效性，并模拟研究了几类多物理场流体耦合模型的应用问题。

 

不可压缩MHD流动的增广Lagrangian块预处理方法

李凌霄

不可压缩磁流体(MHD)方程组有限元离散后会产生一个复杂的双鞍点问题，为了提高计算效率必须发展相应代数方程的高效的预处理算法。报告首先介绍块预处理方法的一些概念，然后从约束优化问题的增广Lagrangian思想出发探讨了Stokes方程、Navier-Stokes方程的一类预处理方法的思路，最后针对MHD方程发展了有效的块预处理方法。此外报告也给出了一些数值算例来验证方法的稳健性和有效性。

 

Network Slimming via Nonconvex Optimization

杨翔宇

The huge size of deep neural networks makes it diffificult to deploy on the embedded platforms with limited computation resources directly. In this talk, we present a novel trimming approach to determine the redundant parameters of the trained deep neural network in a layer-wise manner to produce a compact neural network. This is achieved by minimizing a nonconvex sparsity-inducing term of the network parameters while maintaining the response close to the original one. We present a proximal iteratively reweighted method to resolve the resulting nonconvex model, which approximates the nonconvex objective by a weighted L1 norm of the network parameters. Moreover, to alleviate the computational burden, we develop a novel termination criterion during the subproblem solution, signifificantly reducing the total pruning time. Global convergence analysis and a worst-case O(1/k) ergodic convergence rate for our proposed algorithm is established. Numerical experiments demonstrate the proposed approach is effificient and reliable.

 

小样本环境下多模态数据建模及应用

杨晓慧

本报告面向实际问题，探讨在小样本环境下挖掘多模态数据中蕴含的有用信息，构建可解释性好的深层表示学习模型、多模态关联学习模型和未标记数据驱动的稀疏表示分类模型，进而自主研发硬件要求和成本低，开放兼容的高精度智能化算法及系统。

 

基于余弦边界损失函数的人脸识别

郑 晨

本报告介绍基于深度学习的人脸识别方法中损失函数的发展历程，并在此基础上，针对训练数据可能存在的标注错误问题，提出了一种基于余弦边界损失函数的深度人脸识别模型。该模型通过将标注数据的划分，区分了不同样本的作用。然后，通过一种新的自适应加权的分段损失函数，实现对错误标注样本的抑制。

 

我国建筑业高质量发展的经济效应统计测度——基于CGE模型的分析

苏丽敏

建筑业产业链条长、波及范围广，是我国国民经济支柱产业，研究其高质量发展对新时代我国经济发展的影响具有重要意义。本文在测算建筑业高质量发展水平的基础上，立足系统发展观，构建建筑业可计算一般均衡模型，模拟测度建筑业高质量发展的国民经济效应。研究结果发现：建筑业高质量发展能够有效发挥稳增长、保就业的功效，同时具有较强的出口替代效应；建筑业高质量发展优化了社会资源配置，促进了先进制造业的发展，抑制了非金属矿物制品部门的发展。

 

基于标签一致性的深层半监督非负矩阵分解模型及应用

杨利军

为充分利用数据中携带的标签信息，提升模型的特征表示能力。本报告介绍一种半监督非负矩阵分解模型，该模型是对传统的非负矩阵分解模型的改进。此外，为了降低所提模型对初值的敏感性、挖掘数据的深层特征，进一步构建了基于标签一致性的深层半监督非负矩阵分解模型。多个数据集上的实验结果验证了所提两个模型的性能。

 

Robust Fused Lasso Penalized Huber Regression

辛欣

For some special data in reality, such as the genetic data, adjacent genes may have the similar function. Thus ensuring the smoothness between adjacent genes is highly necessary. But, in this case, the standard lasso penalty just doesn’t seem appropriate anymore. On the other hand, in high-dimensional statistics, some datasets are easily contaminated by outliers or contain variables with heavy-tailed distributions, which makes many conventional methods inadequate. To address both issues, in this paper, we propose an adaptive Huber regression for robust estimation and inference, in which, the fused lasso penalty is used to encourage the sparsity of the coefficients as well as the sparsity of their differences, i.e., local constancy of the coefficient profile. Theoretically, we establish its nonasymptotic estimation error bounds under l2-norm in high-dimensional setting. The proposed estimation method is formulated as a convex, nonsmooth and separable optimization problem, hence, the alternating direction method of multipliers can be employed. In the end, we perform on simulation studies and real cancer data studies, which illustrate that the proposed estimation method is more robust and predictive.

 

An Efficient Method for Joint Delay-Doppler Estimation of Moving Targets in Passive Radar

肖运海

Passive radar systems can detect and track the moving targets of interest by exploiting non-cooperative illuminators-of-opportunity to transmit orthogonal frequency division multiplexing (OFDM) signals. These targets are searched by using a bank of correlators which are tuned to the waveform corresponding to the given Doppler frequency shift and delay. In this talk, we study the problem of joint delay-Doppler estimation of moving targets in OFDM passive radar. This task of estimation is described as an atomic-norm regularized convex optimization problem, and then to a semi-definite programming problem. We show that the directly-extended alternating direction method of multipliers (ADMM) computes each variable in a Gauss-Seidel manner, but its convergence is lack of certificate. To remedy this defect, we employ a symmetric Gauss-Seidel (sGS) into the framework of ADMM, which only needs to compute parts of the subproblems twice but has the ability to ensure the algorithm's convergence. We also do some numerical experiments using some synthetic data which illustrates that the sGS-ADMM is superior to ADMM in terms of the accuracy and computing time. This is a joint work with my coauthors Peili Li and Mengjiao Shi.

 

Adaptive Weighted Curvature-Based Active Contour for Medical Image Segmentation

庞志峰

Image segmentation is a complex and core technique for disease diagnosis or image-guided surgery in the medical image domain. However, low-quality images, such as images with weak edges and intensity inhomogeneities, may bring considerable challenges for radiologists. In this paper, we propose an adaptive weighted curvature-based active contour model by coupling heat kernel convolution and adaptively weighted high-order total variation for medical image segmentation to improve diagnosis effectiveness. To reduce the computational complexity, the heat kernel convolution operation is applied to approximate the perimeter of a segmentation curve. In addition, the weighted parameter included in the high-order total variation term can be automatically evaluated based on an adaptive input image to emphasize local structures and increase segmentation accuracy. Since the proposed method is a smoothing optimization model, the alternating direction method of multipliers is introduced to translate the original problems into several easily solvable subproblems. The numerical experimental results on ultrasonic and MRI datasets demonstrate that the proposed model is quite effificient and robust compared with several traditional segmentation methods. This work is joint with Mengxiao Geng (HENU), Dong Liang(SIAT-CAS) , Zhanli Hu(SIAT-CAS) and Yongming Dai(UIH), Tieyong Zeng(CUHK).