FeatureDistance
FeatureDistance[example1,example2,extractor]
给出由 extractor 定义的特征空间中 example1 和 example2 之间的距离.
FeatureDistance[extractor]
表示可应用于一对范例的 FeatureDistance 的运算符形式.
更多信息和选项
- FeatureDistance 可用于多种数据类型，包括数值、文本、声音和图像以及这些类型的组合等.
- extractor 通常是由 FeatureExtraction 生成的 FeatureExtractorFunction 对象.
- FeatureDistance[extractor][ex1,ex2] 等价于 FeatureDistance[ex1,ex2,extractor].
- 可以给出以下选项：
-
DistanceFunction Automatic 用于特征空间的距离 - 选项 DistanceFunction 可以是任何距离或相异度函数，或定义两个值之间距离的函数 f.
- 对于数值特征向量，缺省设置为 DistanceFunctionEuclideanDistance.
范例
打开所有单元 关闭所有单元基本范例 (1)
在一个简单的数据集上训练 FeatureExtractorFunction，以便与 FeatureDistance 配合使用：
fe = FeatureExtraction[{{1.4, "A"}, {1.5, "A"}, {2.3, "B"}, {5.4, "B"}}]FeatureDistance[{1.8, "A"}, {4.5, "B"}, fe]范围 (1)
在一组颜色上训练 FeatureDistance 函数：
list = ColorData["AlpineColors"] /@ Range[0, 1, 0.1]fe = FeatureExtraction[list]fd = FeatureDistance[fe]fd[Yellow, Orange]根据此距离度量，找出最接近 Red 的颜色：
Nearest[list, Red, 1, DistanceFunction -> fd]选项 (1)
DistanceFunction (1)
fe = FeatureExtraction[{"the cat is grey", "my cat is fast", "this dog is scary", "the big dog"}, "TFIDF"]fd1 = FeatureDistance[fe]fd1["the cat is grey", "the big dog"]距离函数 EuclideanDistance 已被选定. 使用选项 DistanceFunction 来选择 CosineDistance：
fd2 = FeatureDistance[fe, DistanceFunction -> CosineDistance]fd2["the cat is grey", "the big dog"]应用 (1)
dataset = {[image], [image], [image], [image], [image], [image], [image], [image], [image], [image], [image], [image], [image], [image], [image], [image], [image], [image], [image], [image], [image]};fe = FeatureExtraction[dataset]使用 FeatureDistance 计算两个图片之间的特征空间：
FeatureDistance[[image], [image], fe]与提取出的特征之间的 EuclideanDistance 进行比较：
EuclideanDistance@@fe[{[image], [image]}]生成 FeatureDistance 的运算符形式：
fd = FeatureDistance[fe]fd[[image], [image]]dm = DistanceMatrix[dataset, DistanceFunction -> fd];dm // MatrixPlotdataset[[Ordering[Mean[dm], -1]]]属性和关系 (1)
FeatureNearest 可用于在特征空间中寻找相近的样例：
lines = ExampleData[{"Text", "OriginOfSpecies"}, "Lines"];
fe = FeatureExtraction[ExampleData[{"Text", "OriginOfSpecies"}, "Lines"], {"LowerCasedText", "TFIDF", "DimensionReducedVector"}];
FeatureNearest[ExampleData[{"Text", "OriginOfSpecies"}, "Lines"],
"domesticated pigeons", FeatureExtractor -> fe]//Short//AbsoluteTiming这通常比使用带有自定义 DistanceFunction 的 Nearest 更快：
fd = FeatureDistance[fe];
Nearest[lines, "domesticated pigeons", 1, DistanceFunction -> fd]//Short//AbsoluteTiming巧妙范例 (1)
生成手写数字的 NearestNeighborGraph：
mnist = ExampleData[{"MachineLearning", "MNIST"}, "Data"];
fe = FeatureExtraction[Keys[mnist]];
fd = FeatureDistance[fe]digits = Catenate@GroupBy[mnist, Values -> Keys, RandomSample[#, 2]&]NearestNeighborGraph[digits, DistanceFunction -> fd, VertexLabels -> Automatic]文本
Wolfram Research (2016)，FeatureDistance，Wolfram 语言函数，https://reference.wolfram.com/language/ref/FeatureDistance.html.
CMS
Wolfram 语言. 2016. "FeatureDistance." Wolfram 语言与系统参考资料中心. Wolfram Research. https://reference.wolfram.com/language/ref/FeatureDistance.html.
APA
Wolfram 语言. (2016). FeatureDistance. Wolfram 语言与系统参考资料中心. 追溯自 https://reference.wolfram.com/language/ref/FeatureDistance.html 年