发现国内一个学者在同一个会议上发论文10多篇，其中有些有问题，怎么报告？# WaterWorld - 未名水世界
j*m
1 楼
看到有人指出在icmlc 2011会议中同一个作者发表论文达10多篇。本文深知做学术
的不易，觉得在短时间内能在同一个会议上发表10多篇论文，不是造假抄袭，就是粗制
滥造。于是随便找了一篇搜索了一下。发现不如所料，亩产万斤果然是有问题的。
请相关专业人员鉴定是否属于抄袭，以及是否还有其他未发现之处。
抄袭文（以下简称USOM文）
Le Li, Xiaohang Zhang, Zhiwen yu, Zijian Feng, Ruiping Wei, USOM: Mining
and Visualizing Uncertain Data Based on Self-Organizing Maps, Proceedings
of the 2011 International Conference on machine Learning and Cybernetics,
804-809
作者单位：
School of Computer Science and Engineering, South China University of
Technology, Guangzhou, China
Department of Computing, Hong Kong Polytechnic University, Hong Kong
USOM文共6页，但除去参考文献、摘要、致谢之外，还剩4页纸。这4页中，非常明
显与其他3篇未标注论文相同的地方约2页，抄袭部分超过50%。这还只是笔者粗略的发
现，不排除还有其他抄袭的地方未发现。
==========
重复点1：
USOM文：
We represent the fuzzy object o in an uncertain database UD as a
probability distribution function, such as Gaussian distribution, Uniform
distribution, or Poisson distribution. Let o ∈ UD ? Rm be an uncertain item
from UD which can appear at any position x with respect to a probability
density function p(x). Then, the following condition holds:
公式（1）
where Rm is the data space.
被抄文1：
The fuzzy object o in an uncertain database D is represented by a
probability distribution function, such as Gaussian distribution function or
Poisson distribution function. Let o ∈ D ? Rm be an uncertain data item
from D which can appear at any position x with respect to a probability
density function p(x). Then, the following condition holds:
公式（1）
where Rm is the data space.
其中，USOM文的公式（1）与被抄文1的公式（1）一模一样。
==========
重复点2：
USOM文：
The similarity between traditional objects is measured by the distance
between two objects which is represented as a numerical value. Unfortunately
, the similarity of two fuzzy objects cannot be measured by a single value.
As a result, Kriegel H.P.et.al ([7] ) proposed a distance density function.
Let ￠ : UD x UD → R be a metric function(Here we select Euclidean distance
as the distance metric ￠). [嵌入公式] denotes the probability that ￠(o, o
) is not greater than E(where o and o denotes the fuzzy object of and of, E
is a distance threshold). The distance distribution function is defined as
公式（2）
被抄文1：
The similarity between traditional objects is measured by the distance
between two objects which is represented as a numerical value. Unfortunately
, the similarity of two fuzzy objects cannot be measured by a single value.
As a result, Kriegel H.P.et.al ([6]) proposed a distance density function.
Let θ : DxD → R be a metric function(Here we select Euclidean distance as
the distance metric θ). [嵌入公式] denotes the probability that θ(o, o) is
not greater than (where o and o denotes the object o and o, is a distance
threshold). The distance distribution function is defined as
公式（2）
其中，USOM文的公式（2）与被抄文1的公式（2）一模一样。
==========
重复点3：
USOM文：
where p is a probability density function. If the distance ￠(o, o)
between two fuzzy objects is deterministic, the probability density function
is equal to the Diracdelta function δ.
公式（3）
The similarity between the fuzzy objects can be measured by the distance
expectation value E which is the average distance between the fuzzy objects.
公式（4）
被抄文1：
where p is a probability density function. If the distance θ(o, o)
between two objects is deterministic, the probability density function is
equal to the Dirac-delta function δ.
公式（3）
The similarity between the fuzzy objects can be measured by the distance
expectation value Eθ which is the average distance between the fuzzy
objects.
公式（4）
其中，USOM文的公式（3）与被抄文1的公式（3）一样。USOM文的公式（4）与被抄
文1的公式（4）一样。
==========
重复点4：
USOM文的公式（5）与被抄文第2页中2) Competition公式一样。
USOM文：
In the third step, the weight vector u in the wimler neuron n and the
weight vectors of its corresponding neighborhood are updated by
公式（6）
公式（7）
公式（8）（Gaussian Distribution）
where i denotes the ith iteration which corresponds to the ith input
fuzzy object. a(i) is the learning rate.
被抄文2：
An example of neighbourhood function often used is the Gaussian
neighbourhood function,
公式 3)
where a(k) is a monotonically decreasing learning factor at time k.
The winning neuron and the neighbours are adjusted with the rule given
below:
公式 4)
其中，USOM文的公式（6）与被抄文2的公式 4）一样。USOM文的公式（8）与被抄
文2的公式 3) 一样。
==========
重复点5：
USOM文：
Unfortunately, the expected distance E is difficult to compute directly
since the probability density functions (pdf) of the fuzzy objects are
different (e.g. Uniform pdf vs Gaussian pdf). So we substitute s samples for
the fuzzy object. Then, the distance between the fuzzy objects o and u is
redefined as the average distance d(o, u) between the samples of the fuzzy
object o and the weight vector u.
公式（9）
被抄文1：
Unfortunately, the kth nearest neighbor of oi is difficult to compute
since calculating E is time consuming. In order to make FDIC algorithm more
efficient, we substitute s samples for the fuzzy object o. The samples are
selected according to the probability density function p(x) of oi. A minimum
bounding box o.MBR is applied to contain the s samples. Then, the distance
between the fuzzy object o and o is redefined as the average distance d(o, o
) between the samples of the fuzzy object o and o.
公式（16）
其中，USOM文的公式（9）与被抄文1的公式（16）一样。
==========
重复点6：
USOM文：
From the view of the graph theory, the neurons are the nodes in the
graph, while the similarity metrics between the neurons and their neighbor
are the weights of the edges between two nodes. A minimal spanning tree(MST)
connects all the neurons with minimal total weights of the edges. The
weights of the edges are the Mahalanobis distance (d(ni, nj) between the two
neurons ni and nj.
公式 （14）
where ui, uj, Σi, Σj are the weight vectors and variances of the
neurons ni and nj.
MST is constructed by Kruskal's or Prim's algorithm. The desired number
of clusters is obtained by removing the edges in MST whose weights are
greater than a threshold.
被抄文3：
These clusters are viewed as nodes in a graph. A minimal spanning tree(
MST) connects all the clusters with minimal total weights of the edges. The
weights of the edges are the Mahalanobis distance (d(ci, cj)) between the
two clusters ci and cj .
公式（5）
where ui, uj, Σi, Σj are the means and variances of the clusters ci
and cj .
MST is constructed by Kruskal's or Prim's algorithm. The desired number
of clusters is obtained by removing the edges in MST whose weights are
greater than a threshold.
其中，USOM文的公式（14）与被抄文3的公式（5）一样。
==========
重复点7：
USOM文：
All the experiments presented are executed with a Pentium 3.2 GHz CPU
with 1 GByte memory.
The synthetic dataset is created to simulate sensor environment The
synthetic dataset is generated in two stages. In the first stage, we
generate the deterministic dataset D with 500 objects in the [0, 10000]3
unit space by a Gaussian random variable generator.
被抄文3：
All the experiments presented are executed with a Pentium 2.8 GHz CPU
with 1 GByte memory.
We generate two synthetic datasets (10000 points) with Gaussian
distributions in 3D unit space [0, 10000]3 by a Gaussian random variable
generator.
附：
被抄文1：Mining Uncertain Data in Low-dimensional Subspace
被抄文2：Initialization of Self-Organizing Maps: Principal Components
Versus Random Initialization. A Case Study
被抄文3：GCA: A real-time grid-based clustering algorithm for large data
set
的不易，觉得在短时间内能在同一个会议上发表10多篇论文，不是造假抄袭，就是粗制
滥造。于是随便找了一篇搜索了一下。发现不如所料，亩产万斤果然是有问题的。
请相关专业人员鉴定是否属于抄袭，以及是否还有其他未发现之处。
抄袭文（以下简称USOM文）
Le Li, Xiaohang Zhang, Zhiwen yu, Zijian Feng, Ruiping Wei, USOM: Mining
and Visualizing Uncertain Data Based on Self-Organizing Maps, Proceedings
of the 2011 International Conference on machine Learning and Cybernetics,
804-809
作者单位：
School of Computer Science and Engineering, South China University of
Technology, Guangzhou, China
Department of Computing, Hong Kong Polytechnic University, Hong Kong
USOM文共6页，但除去参考文献、摘要、致谢之外，还剩4页纸。这4页中，非常明
显与其他3篇未标注论文相同的地方约2页，抄袭部分超过50%。这还只是笔者粗略的发
现，不排除还有其他抄袭的地方未发现。
==========
重复点1：
USOM文：
We represent the fuzzy object o in an uncertain database UD as a
probability distribution function, such as Gaussian distribution, Uniform
distribution, or Poisson distribution. Let o ∈ UD ? Rm be an uncertain item
from UD which can appear at any position x with respect to a probability
density function p(x). Then, the following condition holds:
公式（1）
where Rm is the data space.
被抄文1：
The fuzzy object o in an uncertain database D is represented by a
probability distribution function, such as Gaussian distribution function or
Poisson distribution function. Let o ∈ D ? Rm be an uncertain data item
from D which can appear at any position x with respect to a probability
density function p(x). Then, the following condition holds:
公式（1）
where Rm is the data space.
其中，USOM文的公式（1）与被抄文1的公式（1）一模一样。
==========
重复点2：
USOM文：
The similarity between traditional objects is measured by the distance
between two objects which is represented as a numerical value. Unfortunately
, the similarity of two fuzzy objects cannot be measured by a single value.
As a result, Kriegel H.P.et.al ([7] ) proposed a distance density function.
Let ￠ : UD x UD → R be a metric function(Here we select Euclidean distance
as the distance metric ￠). [嵌入公式] denotes the probability that ￠(o, o
) is not greater than E(where o and o denotes the fuzzy object of and of, E
is a distance threshold). The distance distribution function is defined as
公式（2）
被抄文1：
The similarity between traditional objects is measured by the distance
between two objects which is represented as a numerical value. Unfortunately
, the similarity of two fuzzy objects cannot be measured by a single value.
As a result, Kriegel H.P.et.al ([6]) proposed a distance density function.
Let θ : DxD → R be a metric function(Here we select Euclidean distance as
the distance metric θ). [嵌入公式] denotes the probability that θ(o, o) is
not greater than (where o and o denotes the object o and o, is a distance
threshold). The distance distribution function is defined as
公式（2）
其中，USOM文的公式（2）与被抄文1的公式（2）一模一样。
==========
重复点3：
USOM文：
where p is a probability density function. If the distance ￠(o, o)
between two fuzzy objects is deterministic, the probability density function
is equal to the Diracdelta function δ.
公式（3）
The similarity between the fuzzy objects can be measured by the distance
expectation value E which is the average distance between the fuzzy objects.
公式（4）
被抄文1：
where p is a probability density function. If the distance θ(o, o)
between two objects is deterministic, the probability density function is
equal to the Dirac-delta function δ.
公式（3）
The similarity between the fuzzy objects can be measured by the distance
expectation value Eθ which is the average distance between the fuzzy
objects.
公式（4）
其中，USOM文的公式（3）与被抄文1的公式（3）一样。USOM文的公式（4）与被抄
文1的公式（4）一样。
==========
重复点4：
USOM文的公式（5）与被抄文第2页中2) Competition公式一样。
USOM文：
In the third step, the weight vector u in the wimler neuron n and the
weight vectors of its corresponding neighborhood are updated by
公式（6）
公式（7）
公式（8）（Gaussian Distribution）
where i denotes the ith iteration which corresponds to the ith input
fuzzy object. a(i) is the learning rate.
被抄文2：
An example of neighbourhood function often used is the Gaussian
neighbourhood function,
公式 3)
where a(k) is a monotonically decreasing learning factor at time k.
The winning neuron and the neighbours are adjusted with the rule given
below:
公式 4)
其中，USOM文的公式（6）与被抄文2的公式 4）一样。USOM文的公式（8）与被抄
文2的公式 3) 一样。
==========
重复点5：
USOM文：
Unfortunately, the expected distance E is difficult to compute directly
since the probability density functions (pdf) of the fuzzy objects are
different (e.g. Uniform pdf vs Gaussian pdf). So we substitute s samples for
the fuzzy object. Then, the distance between the fuzzy objects o and u is
redefined as the average distance d(o, u) between the samples of the fuzzy
object o and the weight vector u.
公式（9）
被抄文1：
Unfortunately, the kth nearest neighbor of oi is difficult to compute
since calculating E is time consuming. In order to make FDIC algorithm more
efficient, we substitute s samples for the fuzzy object o. The samples are
selected according to the probability density function p(x) of oi. A minimum
bounding box o.MBR is applied to contain the s samples. Then, the distance
between the fuzzy object o and o is redefined as the average distance d(o, o
) between the samples of the fuzzy object o and o.
公式（16）
其中，USOM文的公式（9）与被抄文1的公式（16）一样。
==========
重复点6：
USOM文：
From the view of the graph theory, the neurons are the nodes in the
graph, while the similarity metrics between the neurons and their neighbor
are the weights of the edges between two nodes. A minimal spanning tree(MST)
connects all the neurons with minimal total weights of the edges. The
weights of the edges are the Mahalanobis distance (d(ni, nj) between the two
neurons ni and nj.
公式 （14）
where ui, uj, Σi, Σj are the weight vectors and variances of the
neurons ni and nj.
MST is constructed by Kruskal's or Prim's algorithm. The desired number
of clusters is obtained by removing the edges in MST whose weights are
greater than a threshold.
被抄文3：
These clusters are viewed as nodes in a graph. A minimal spanning tree(
MST) connects all the clusters with minimal total weights of the edges. The
weights of the edges are the Mahalanobis distance (d(ci, cj)) between the
two clusters ci and cj .
公式（5）
where ui, uj, Σi, Σj are the means and variances of the clusters ci
and cj .
MST is constructed by Kruskal's or Prim's algorithm. The desired number
of clusters is obtained by removing the edges in MST whose weights are
greater than a threshold.
其中，USOM文的公式（14）与被抄文3的公式（5）一样。
==========
重复点7：
USOM文：
All the experiments presented are executed with a Pentium 3.2 GHz CPU
with 1 GByte memory.
The synthetic dataset is created to simulate sensor environment The
synthetic dataset is generated in two stages. In the first stage, we
generate the deterministic dataset D with 500 objects in the [0, 10000]3
unit space by a Gaussian random variable generator.
被抄文3：
All the experiments presented are executed with a Pentium 2.8 GHz CPU
with 1 GByte memory.
We generate two synthetic datasets (10000 points) with Gaussian
distributions in 3D unit space [0, 10000]3 by a Gaussian random variable
generator.
附：
被抄文1：Mining Uncertain Data in Low-dimensional Subspace
被抄文2：Initialization of Self-Organizing Maps: Principal Components
Versus Random Initialization. A Case Study
被抄文3：GCA: A real-time grid-based clustering algorithm for large data
set
