A Machine Learning Approach to Automated Negotiation

Automated negotiation between two competitive agents is analyzed, and a multi-issue negotiation model based on machine learning, time belief, offer belief and state-action pair expected Q value is developed. Unlike the widely used approaches such as game theory approach, heuristic approach and argumentation approach, This paper uses a machine learning method to compute agents‘ average Q values in each negotiation stage. The delayed reward is used to generate agents‘ offer and counteroffer of every issue. The effect of time and discount rate on negotiation outcome is analyzed. Theory analysis and experimental data show this negotiation model is practical.     

Prediction of Tumor Outcome Based on Gene Expression Data

Gene expression mieroarray data can be used to classify tumor types.We proposed a new procedure to classify human tumor samples based on mieroarray gene expressions by using a hybrid supervised learning method called MOEA WV(Multi-Objective Evolutionary Algorithm Weighted Voting).MOEA is used to search for a relatively few subsets of informative genes from the high-dimensional gene space,and WV is used as a classification tool.This new method has been applied to predicate the subtypes of lymphoma and outcomes of medulloblastoma.The results are relatively accurate and meaningful compared to those from other methods.     

Biological networks to the analysis of microarray data

Microarray technology, which permits rapid and large-scale screening for patterns of gene expressions, usually generates a large amount of data. How to mine the biological meanings under these data is one of the main challenges in bioinformatics. Compared to the pure mathematical techniques, those methods incorporated with some prior biological knowledge generally bring better interpretations. Recently, a new analysis, in which the knowledge of biological networks such as metabolic network and protein interaction network is introduced, is widely applied to microarray data analysis. The microarray data analysis based on biological networks contains two main research aspects: identification of active components in biological networks and assessment of gene sets significance. In this paper, we briefly review the progress of these two categories of analyses, especially some representative methods.     

Improvement of the classification accuracy in discriminating diabetic retinopathy by multifocal electroretinogram analysis

The multifocal electroretinogram (mfERG) is a newly developed electrophysiological technique. In this paper, a classification method is proposed for early diagnosis of the diabetic retinopathy using mfERG data. MfERG records were obtained from eyes of healthy individuals and patients with diabetes at different stages. For each mfERG record, 103 local responses were extracted. Amplitude value of each point on all the mfERG local responses was looked as one potential feature to classify the experimental subjects. Feature subsets were selected from the feature space by comparing the inter-intra distance. Based on the selected feature subset, Fisher’s linear classifiers were trained. And the final classification decision of the record was made by voting all the classifiers’ outputs. Applying the method to classify all experimental subjects, very low error rates were achieved. Some crucial properties of the diabetic retinopathy classification method are also discussed.     

Mining and Integrating Reliable Decision Rules for Imbalanced Cancer Gene Expression Data Sets

There have been many skewed cancer gene expression datasets in the post-genomic era. Extraction of differential expression genes or construction of decision rules using these skewed datasets by traditional algorithms will seriously underestimate the performance of the minority class, leading to inaccurate diagnosis in clinical trails. This paper presents a skewed gene selection algorithm that introduces a weighted metric into the gene selection procedure. The extracted genes are paired as decision rules to distinguish both classes, with these decision rules then integrated into an ensemble learning framework by majority voting to recognize test examples; thus avoiding tedious data normalization and classifier construction. The mining and integrating of a few reliable decision rules gave higher or at least comparable classification performance than many traditional class imbalance learning algorithms on four benchmark imbalanced cancer gene expression datasets.     

New feature extraction in gene expression data for tumor classification

Using gene expression data to discriminate tumor from the normal ones is a powerful method. However, it is sometimes difficult because the gene expression data are in high dimension and the object number of the data sets is very small. The key technique is to find a new gene expression profiling that can provide understanding and insight into tumor related cellular processes. In this paper, we propose a new feature extraction method based on variance to the center of the class and employ the support vector machine to recognize the gene data either normal or tumor. Two tumor data sets are used to demonstrate the effectiveness of our methods. The results show that the performance has been significantly improved.     

A Modified Extended Bayesian Method for Parameter Estimation

This paper presents a modified extended Bayesian method for parameter estimation. In this method the mean value of the a priori estimation is taken from the values of the estimated parameters in the previous iteration step. In this way, the parameter covariance matrix can be automatically updated during the estimation procedure, thereby avoiding the selection of an empirical parameter. Because the extended Bayesian method can be regarded as a Tikhonov regularization, this new method is more stable than both the least-squares method and the maximum likelihood method. The validity of the proposed method is illustrated by two examples: one based on simulated data and one based on real engineering data.     

A GENERALLY APPLICABLE METHOD OF STATISTIC ANALYSIS FOR STUDYING AND CLASSIFYING NERVE TYPES OF CHILDREN,JUVENILES AND YOUTHS

In this paper, a method of statistic analysis is suggested for studying and classifying nerve types of children, juveniles and youths. First, a set of effective calculating formulae is established so as to describe nerve types quantitatively. Thus, our analysis of nerve types is based on the experimental data. Since these data can not be considered as normally distributed, even the sampling is sized greatly. This has caused certain difficulties in the data analysis. The Box-Cox Transformation is proposed for the sampling data in this paper. As a result, direr tronsformation all the sets of experimental data can successfully pass Kolmogorov‘s no rmallsation test. And then, the suitable representative points in normal distribution are sleeted as classifying standards. Once the dividing points for classification are set up, the problem of classifying nerve types is solved successfully.     

A fast SVM training algorithm based on the set segmentation and k-means clustering

At present, studies on training algorithms for support vector machines (SVM) are important issues in the field of machine learning. It is a challenging task to improve the efficiency of the algorithm without reducing the generalization performance of SVM. To face this challenge, a new SVM training algorithm based on the set segmentation and k-means clustering is presented in this paper. The new idea is to divide all the original training data into many subsets, followed by clustering each subset using k-means clustering and finally train SVM using the new data set obtained from clustering centroids. Considering that the decomposition algorithm such as SVMlight is one of the major methods for solving support vector machines, the SVMlight is used in our experiments. Simulations on different types of problems show that the proposed method can solve efficiently not only large linear classification problems but also large nonlinear ones.     

Big data analysis of the human brain’s functional interactions based on fMRI

The human brain is a huge, complex system generating brain activity. The exploration of human brain function using functional magnetic resonance imaging （fMRI） is a promising method to understand brain activity. However, the complexity of the big data generated by fMRI facilitates the analysis of various levels of human brain activity, such as the distribution of neural represen- tations, the interaction between different regions, and the dynamic interaction over time. These different levels can depict distinct prospects of the human brain activity, and considerable progress has been achieved. In the future, more big data analysis methods combining advances in computer science, including larger-scale computing, machine learning, and graph theory, will promote the understanding of the human brain.     

Gene expression profiling predicts clinical outcome of breast cancer

Breast cancer patients with the same stage of disease can have markedly different treatment responses and overall outcome. The strongest predictors for metastases (for example, lymph node status and histological grade) fail to classify accurately breast tumours according to their clinical behaviour. Chemotherapy or hormonal therapy reduces the risk of distant metastases by approximately one-third; however, 70-80% of patients receiving this treatment would have survived without it. None of the signatures of breast cancer gene expression reported to date allow for patient-tailored therapy strategies. Here we used DNA microarray analysis on primary breast tumours of 117 young patients, and applied supervised classification to identify a gene expression signature strongly predictive of a short interval to distant metastases ('poor prognosis' signature) in patients without tumour cells in local lymph nodes at diagnosis (lymph node negative). In addition, we established a signature that identifies tumours of BRCA1 carriers. The poor prognosis signature consists of genes regulating cell cycle, invasion, metastasis and angiogenesis. This gene expression profile will outperform all currently used clinical parameters in predicting disease outcome. Our findings provide a strategy to select patients who would benefit from adjuvant therapy.     

特征提取与多目标机器学习研究及应用

特征提取与多目标机器学习算法是基于多目标协同EA提出的,该算法通过对学习样本多属性进行特征提取找出其核属性,由核属性与其他非核属性组成属性组,从而提高了分类的精度。各属性组再按相似性和类标签进行有监督地聚成类簇,类簇个数和中心根据适应度矢量函数通过机器学习算法自动确定,这样类簇个数和中心就不受主观因素的影响并且保证了这两个关键要素的优化性质。待分类样本的类属是按离某个类簇中心距离最近邻法则和该类簇的类标签来判定。最后,将算法应用到UCI数据集中的Liver Disorders和Hepatitis两个数据集,以及浙江省北部地区夏天异常高温天气预测。通过实验表明,特征提取与机器学习算法优于著名的朴素贝叶斯、C4.5、SVM算法。     

结合先验信息的多属性诊断测验分类研究

在动态的学习过程中,随着学习的深入和知识属性的个数逐渐增加,学生的知识状态也会发生动态变化.在这样的应用场景下如何结合先验信息提高诊断测验的判准率具有较大的挑战.该文提出基于学生对已学习属性的掌握概率来预测包含新属性后的属性向量的先验信息.考虑2种实际的应用情境,并通过模拟和实证研究来评价该方法的表现.结果表明：在2种实际学习情境中,结合先验信息在多属性诊断测验中能起到提高判准率的作用,其中在基于个体先验信息时的表现更好.基于预测先验信息的方法也摆脱了在以往研究中不同学习阶段测验属性个数一致或数量的限制,使得分类精度有较大改进.实证数据分析进一步表明该方法具有较高的应用价值.     

结合主题模型与自监督学习的可控文本生成技术研究

基于大型预训练语言模型的有监督学习方法在可控文本生成任务上取得了优秀的成果，但这些研究都着重于控制生成文本的高级属性(比如情感与主题),而忽略了泛化性问题.现有的基于自监督学习的研究方法则通过句子级别的训练来使模型获得补全整句的能力，使模型做到单词和短语级别的控制生成，但生成与特定属性强相关句子的能力依旧待提升.所以本文提出了一种单词级别(细粒度)与句子(粗粒度)级别相结合的多粒度训练方式：单词级别的主题模型让模型学习主题层面的语义以获得主题到文本的生成能力，句子级别的自监督训练让模型学习整句的表征以获得补全句子的能力.通过主题模型与自监督学习的结合，使模型在单词与短语级别的可控生成阶段取得了更好的效果.实验表明，本文提出的模型在主题契合度以及常规文本生成指标方面优于现有的基线模型.     

基于加权极限学习机的肿瘤基因表达谱数据分类

基因表达谱数据一般来源于临床试验,而在临床试验中,试验样本的类分布情况是不确定的,这就使得表达谱数据往往具有比较明显的不平衡性.采用加权极限学习机来对不平衡基因表达谱数据进行分类,为了减少因为不平衡数据引起的分类误差,一个临时的权重被分配给每一个样本以增强少样本类的影响,同时减少多样本类的影响,进而提高肿瘤分类的准确率.实验结果表明,所提方法能够提高少样本类的识别率,从而提高分类器的总体性能.     

基于模糊核判别分析的基因表达数据分析方法

针对基因表达数据空间分布的特性,提出了一种基于模糊核判别分析的基因表达数据分析方法.方法综合了模糊数学以及核判别分析方法的优点,提高了对基因表达数据分类识别的准确性.以多发性骨髓瘤的基因表达数据为例进行了实验,从实验结果可以看出,采用模糊核判别分析方法可以得到最佳的识别效果.     

基于颜色图直方图的监督颜色恒常性算法

在颜色物体识别等应用中，获取不随光照变化的颜色恒常性特征是问题的关键。监督颜色恒常性方法，即通过在环境中放置校正用的颜色片来计算并消除光照的影响，由于其简单方便而得到了广泛的应用。在本文中，基于对直方图光照不变量的分析，提出了一种新的监督颜色恒常性算法。该算法与传统算法相比较，不需要知道颜色片的反射率和成象系统的通道响应，从而大大简化了实现的要求。文中给出了算法的基本思想，并对实验结果进行了分析。     

基于Kohonen神经网络的遥感影像监督分类

作为一种自组织网络,Kohonen神经网络在遥感影像聚类中得到了广泛的应用.但Kohonen算法具有初始权值随机化、学习率和邻域难以确定等缺陷,并且作为一种聚类算法,难以直接进行监督分类.对Kohonen非监督学习算法进行改进,用最大最小距离法确定初始权值,利用权值误差作为网络训练收敛条件,并将Kohonen算法与奖惩...     

基于人脸运动单元及表情关系模型的自动表情识别

面部表情是人们表达情感和意向最有效、自然、快捷的方式. 表情的发生主体不同,程度不同,导致了表情的自发性. 基于这一难点,建立了一种人脸运动单元(action units,AUs)及面部表情间的概率关系模型,该模型将人脸分为眉眼区域和嘴巴区域两部分,采用Gabor小波提取区域特征,通过K临近(K nearest neighbor,KNN)与贝叶斯网络(Bayesian network,BN)相结合的机器学习算法进行自动AUs表情识别. 这种改进的机器学习算法,通过训练数据以及主观的先验知识进行模型学习,为AUs配以不同的权重,并且根据极大后验概率(maximum a posteriori probability,MAP)选取最优表情. 实验表明,本文所提出的模型对不同主体、不同程度的表情都表现出了较高的识别率,是一种高效且鲁棒性强的自动表情识别系统.