SamSPECTRAL(SamSPECTRAL)
SamSPECTRAL()所属R语言包：SamSPECTRAL
Identifies the cell populations in flow cytometry data.
标识流式单元仪数据的单元群。
译者：生物统计家园网 机器人LoveR
描述----------Description----------
Given an FCS file as input, SamSPECTRAL first builds the communities to sample the data points. Then, it builds a graph and after weighting the edges of the graph by conductance computation, it is passed to a classic spectral clustering algorithm to find the spectral clusters. The last stage of SamSPECTRAL is to combine the spectral clusters. The resulting "connected components" estimate biological cell populations in the data sample.
鉴于FCS的文件作为输入，SamSPECTRAL首次建立了社区采样数据点。然后，它建立一个图和加权后的电导计算图的边缘，它是通过一个经典的谱聚类算法找到谱聚类。最后阶段的SamSPECTRAL结合光谱聚类。由此产生的“连接组件”估计数据样本中的生物单元群。
用法----------Usage----------
SamSPECTRAL(data.points, dimensions=1:dim(data.points)[2], normal.sigma, separation.factor,number.of.clusters = NA, scale=rep(1,dim(data.points)[2]),
talk = TRUE, precision = 6, eigenvalues.num =NA, return_only.labels=TRUE, do.sampling=TRUE, beta=4, stabilizer=1000)
参数----------Arguments----------
参数：data.points
A matrix that contains coordinates of the data points.
矩阵包含的数据点的坐标。
参数：dimensions
A vector that determines which dimension of the data point matrix are chosen for investigation.
一个向量，确定数据点矩阵维度选择的调查。
参数：normal.sigma
A scaling parameter that determines the "resolution" in the spectral clustering stage. By increasing it, more spectral clusters are identified. This can be useful when "small" population are aimed. See the user manual for a suggestion on how to set this parameter using the eigenvalue curve.
缩放参数，决定在谱聚类阶段的“决议”。通过增加，多光谱聚类确定。这可能是有用的，当“小”的人口的目的。如何设置这个参数，使用特征值曲线的建议，请参阅用户手册。
参数：separation.factor
This threshold controls to what extend clusters should be combined or kept separate.Normally, an appropriate value will fall in range 0.3-2.
此阈值控制哪些扩展聚类，应结合或separate.Normally保持一个适当的值将下降0.3-2范围。
参数：number.of.clusters
The default value is NA which leads to computing the number of spectral clusters automatically, otherwise this number will determine the number of spectral clusters.
默认值是NA，从而导致计算光谱聚类的数量，自动，否则这个数字将决定光谱聚类。
参数：talk
A boolean flag with default value TRUE. Setting it to FALSE will keep running the procedure quite with no messages.
一个布尔标志使用默认值true。设置为false，将继续运行程序相当没有消息。
参数：precision
Determines the precision of computations. Setting it to 6 will work and increasing it does not improve results.
决定了计算精度。它设置为6，工作和增加不改善的结果。
参数：eigenvalues.num
An integer with default value NA which prevents ploting the curve of eigenvalues. Otherwise, they will be ploted upto this number.
与防止作图的曲线特征值的默认值的NA整数。否则，他们将ploted高达这个数字。
参数：return_only.labels
A boolean flag with default value TRUE. If the user set it to FALSE, SamSPECTRAL function will return all the intermediate objects that are computed during the sampling, similarity calculation, spectral clustering and combining stages.
一个布尔标志使用默认值true。如果用户设置为false，SamSPECTRAL函数将返回所有的采样，相似度计算，谱聚类相结合的阶段，在计算的中间对象。
参数：do.sampling
A boolean flag with default value TRUE. If set to FALSE, the sampling stage will be ignored by picking up all the data points.
一个布尔标志使用默认值true。如果设置为FALSE，拿起所有的数据点采样阶段将被忽略。
参数：beta
A parameter with default value 4 which must NOT be changed except for huge samples with more than 100,000 data points or for developmental purposes. Setting beta to zero will reduce computational time by applying the following approximation to the conductance calculation step. For each two community, the conductance will be the conductance between their representatives times their sizes.
除了巨大的样品10多万个数据点，或为发展目的，绝不能改变的参数与默认值4。测试设置到零，将采用以下近似的电导计算步骤，减少计算时间。对于每两个社区，电导将是他们的代表之间的电导倍大小。
参数：scale
A vector the length of which is equal to the number of dimensions. The coordinates in each dimension are multiplied by the corresponding scaling factor. So, the bigger this factor is for a dimension, SamSPECTRAL will consider that dimension to be "more significant" and consequently, that dimension will be more effective in clustering.
其中一个向量的长度是相等的维数。在每个维度的坐标乘以相应的比例因子。因此，这个因素是更大的一个维度，SamSPECTRAL将考虑该维度是“更重要的”，因此，该维度的将是更有效的聚类。
参数：stabilizer
The larger this integer is, the final results will be more stable because the underlying kmeans will restart many more times.
这个整数是较大的，最后的结果会更加稳定，因为底层的KMEANS将重新启动很多倍。
Details
详情----------Details----------
Hints for setting separation.factor and normal.sigma: While separation.factor=0.7 is normally an appropriate value for many datasets, for others some value in range 0.3 to 1.2 may produce better results depending on what populations are of particular interest. The larger normal.sigma is the algorithm will find smaller clusters. It can be adjusted best by considering the plot of eigenvalues as explained in the vignette.
提示设置separation.factor和normal.sigma：虽然separation.factor= 0.7对许多数据集通常是一个适当的值，为他人，一些值在0.3至1.2范围可能会产生更好的效果取决于人口特别的兴趣。较大的normal.sigma算法会找到较小的簇。它可以调整最好考虑积的特征值中的小插曲解释。
值----------Value----------
Returns a vector of labels for data points. If the input parameter return_only.labels is set to FALSE, all the objects that are computed during the intermediate will be returned including: society for sampling stage, conductance for similarity calculation, and clustering_result.
返回一个数据点标签的向量。如果输入参数return_only.labels的设置为FALSE，所有被计算在中间的对象包括：社会抽样阶段，为相似度计算的电导和clustering_result的，将返回。
作者（S）----------Author(s)----------
Habil Zare and Parisa Shooshtari
参考文献----------References----------
<h3>See Also</h3> <code>SamSPECTRAL</code>, <code>Building_Communities</code>, <code>Conductance_Calculation</code>, <code>Civilized_Spectral_Clustering</code>, <code>Connecting</code>
举例----------Examples----------
## Not run: [＃无法运行：]
library(SamSPECTRAL)
# Reading data file which has been transformed using log transform[读取数据文件已使用log变换转化]
data(small_data)
full <- small
L <- SamSPECTRAL(data.points=full,dimensions=c(1,2,3), normal.sigma = 200, separation.factor = 0.39)
plot(full, pch='.', col= L)
## End(Not run) [＃结束（不运行）]
转载请注明:出自 生物统计家园网(http://www.biostatistic.net)。
注：
注1：为了方便大家学习，本文档为生物统计家园网机器人LoveR翻译而成，仅供个人R语言学习参考使用，生物统计家园保留版权。
注2：由于是机器人自动翻译，难免有不准确之处，使用时仔细对照中、英文内容进行反复理解，可以帮助R语言的学习。
注3：如遇到不准确之处，请在本贴的后面进行回帖，我们会逐渐进行修订。
|
发表于 2012-2-26 13:50:55
