This paper presents the contribution1 of the datascience Kitchen at GermEval 2021 shared task on the identification of toxic, engaging, and fact-claiming comments. The task aims at extending the identification of off...
详细信息
In this paper, we are concerned with the problem of the pathwise uniqueness of one-dimensional reflected stochastic differential equations with jumps under the assumption of non-Lipschitz continuous coefficients whose...
详细信息
In this paper, we are concerned with the problem of the pathwise uniqueness of one-dimensional reflected stochastic differential equations with jumps under the assumption of non-Lipschitz continuous coefficients whose proof are based on the technique of local time.
This column is about raising questions, rather than providing answers. These days “data based decision making” is the rage among administrators in both industry and academia. The desire for this dependence on algori...
This column is about raising questions, rather than providing answers. These days “data based decision making” is the rage among administrators in both industry and academia. The desire for this dependence on algorithms stems from the general idea that “humans are biased but machines are not”. More and more social decisions, like qualifying for welfare are, are made using algorithms. With this, the data scientists, who are behind the algorithms, are given a lot of power (and responsibility.) In this column, I discuss demographic characteristics of data scientists with conjectures on why this group is non-diverse. Should we allow a small group of non-representative people to make decisions that affect affect larger society?
作者：
Tru CaoDepartment of Biostatistics and Data Science
The University of Texas Health Science Center at Houston School of Public Health Chair: Cuong Pham (Posts & Telecommunications Institute of Technology Vietnam
We are witnessing the beginning of a data driven era with the explosion of data, impact of data on our everyday lives, and advances of data processing methodology and technology. At this juncture, datascience has eme...
We are witnessing the beginning of a data driven era with the explosion of data, impact of data on our everyday lives, and advances of data processing methodology and technology. At this juncture, datascience has emerged as an interdisciplinary field to deal with data for which statistics and machine learning are two key enablers. Originally, statistics and machine learning appear to have been developed in the different contexts of mathematics and computer science, respectively. datascience has brought them together in which statistics focuses on mathematical foundations and methods, while machine learning is more on algorithms and automated data processing. First, this talk looks back on a timeline of the emergence of statistics, machine learning, datascience, and related fields. Second, it reviews a chronology of the invention and context of some important statistical and machine learning methods. Third, it discusses relationships between statistics, machine learning, and datascience. Finally, it addresses some challenges and overcoming ways of learning from big data.
Recently,self-supervised learning has shown great potential in Graph Neural Networks (GNNs) through contrastive learning,which aims to learn discriminative features for each node without label information. The key to ...
详细信息
Recently,self-supervised learning has shown great potential in Graph Neural Networks (GNNs) through contrastive learning,which aims to learn discriminative features for each node without label information. The key to graph contrastive learning is data augmentation. The anchor node regards its augmented samples as positive samples,and the rest of the samples are regarded as negative samples,some of which may be positive samples. We call these mislabeled samples as “false negative” samples,which will seriously affect the final learning effect. Since such semantically similar samples are ubiquitous in the graph,the problem of false negative samples is very significant. To address this issue,the paper proposes a novel model,False negative sample Detection for Graph Contrastive Learning (FD4GCL),which uses attribute and structure-aware to detect false negative samples. Experimental results on seven datasets show that FD4GCL outperforms the state-of-the-art baselines and even exceeds several supervised methods.
With the acceleration of urbanization construction, the contradiction between supply and demand of urban public transportation resources is becoming increasingly prominent, resulting in increasingly serious problems s...
详细信息
For given l,s∈N,Λ={ρ_(j)}_(j=i),…,s,ρj∈T,the C^(*)-algebra B:=ε({r_(j)}_(j=1),…,s,Λ,l)is defined to be the universal C^(*)-algebra generated by l unitaries u_(1),…,u_(l) subject to the relations r_(j)(u_(1),...
详细信息
For given l,s∈N,Λ={ρ_(j)}_(j=i),…,s,ρj∈T,the C^(*)-algebra B:=ε({r_(j)}_(j=1),…,s,Λ,l)is defined to be the universal C^(*)-algebra generated by l unitaries u_(1),…,u_(l) subject to the relations r_(j)(u_(1),…,u_(l))-ρ_(j)=0 for all j=1,…,s,where the r_(j) is monomial in u_(1),…,u_(l) and their inverses for j=1,2,…,*** B is a unital AF-algebra with a unique tracial state,and K_(0)(B)is a finitely generated group,we say that the relations({r_(j)}_(j=1),…,s,Λ,l)are *** the relations({r_(j)}_(j=1),…,s,Λ,l)are AF-relations,we prove that,for any ε>0,there exists a δ>0 satisfying the following:for any unital C^(*)-algebra A with the cancellation property,strict comparison,nonempty tracial state space,and any l unitaries u_(1),u_(2),…,u_(l)∈A satisfying‖r_(j)(u_(1),u_(2),…,u_(l))-ρ_(j)‖<δ,j=1,2,…,s,and certain trace conditions,there exist l unitaries u_(1),u_(2),…,u_(l)∈A such that r_(j)(u_(1),u_(2),…,u_(l))=ρ_(j) for j=1,2,…,s,and‖ui-ui‖<ε for i=1,2,…,***,we give several applications of the above result.
The Polar Environment datascience Center (PEDSC) is one of the centers of the Joint Support-Center for datascience Research (DS) of the Research Organization of Information and Systems (ROIS), which was established ...
详细信息
Classic Graph Convolutional Networks (GCNs) often learn node representation holistically, which ignores the distinct impacts from different neighbors when aggregating their features to update a node’s representation....
详细信息
Classic Graph Convolutional Networks (GCNs) often learn node representation holistically, which ignores the distinct impacts from different neighbors when aggregating their features to update a node’s representation. Disentangled GCNs have been proposed to divide each node’s representation into several feature units. However, current disentangling methods do not try to figure out how many inherent factors the model should assign to help extract the best representation of each node. This paper then proposes D^(2)-GCN to provide dynamic disentanglement in GCNs and present the most appropriate factorization of each node’s mixed features. The convergence of the proposed method is proved both theoretically and experimentally. Experiments on real-world datasets show that D^(2)-GCN outperforms the baseline models concerning node classification results in both single- and multi-label tasks.
暂无评论