Wednesday, September 19, 2012

hausdorff距离 h(A,B)实际上首先对点集A中的每个点ai到距离此点ai最近的B集合中点bj之间的距离‖ai-bj‖进行排序,然后取该距离中的最大值作为h(A,B)的值.h(B,A)同理可得;to think about growth rather than debt.”

描述子距离种类
186人阅读评论(0)收藏举报
1.hausdorff距离
  微分动力系统原理 这本书里有介绍
  Hausdorff距离是描述两组点集之间相似程度的一种量度,它是两个点集之间距离的一种定义形式:假设有两组集合A={a1,…,ap},B={b1,…,bq},则这两个点集合之间的Hausdorff距离定义为H(A,B)=max(h(A,B),h(B,A)) (1)
  其中,
  h(A,B)=max(a∈A)min(b∈B)‖a-b‖ (2)
  h(B,A)=max(b∈B)min(a∈A)‖b-a‖ (3)
  ‖·‖是点集A和B点集间的距离范式(如:L2或Euclidean距离).
  这里,式(1)称为双向Hausdorff距离,是Hausdorff距离的最基本形式;式(2)中的h(A,B)和h(B,A)分别称为从A集合到B集合和从B集合到A集合的单向Hausdorff距离.即h(A,B)实际上首先对点集A中的每个点ai到距离此点ai最近的B集合中点bj之间的距离‖ai-bj‖进行排序,然后取该距离中的最大值作为h(A,B)的值.h(B,A)同理可得.
  由式(1)知,双向Hausdorff距离H(A,B)是单向距离h(A,B)和h(B,A)两者中的较大者,它度量了两个点集间的最大不匹配程度


2.欧式距离


  欧几里得距离定义: 欧几里得距离( Euclidean distance)也称欧式距离,它是一个通常采用的距离定义,它是在m维空间中两个点之间的真实距离。
  在二维和三维空间中的欧式距离的就是两点之间的距离,二维的公式是
  d = sqrt((x1-x2)^+(y1-y2)^)
  三维的公式是
  d=sqrt((x1-x2)^+(y1-y2)^+(z1-z2)^)
  推广到n维空间,欧式距离的公式是
  d=sqrt( ∑(xi1-xi2)^ ) 这里i=1,2..n
  xi1表示第一个点的第i维坐标,xi2表示第二个点的第i维坐标
  n维欧氏空间是一个点集,它的每个点可以表示为(x(1),x(2),...x(n)),其中x(i)(i=1,2...n)是实数,称为x的第i个坐标,两个点x和y=(y(1),y(2)...y(n))之间的距离d(x,y)定义为上面的公式.
  欧氏距离看作信号的相似程度。 距离越近就越相似,就越容易相互干扰,误码率就越高。
  所谓欧氏距离变换,是指对于一张二值图像(再次我们假定白色为前景色,黑色为背景色),将前景中的像素的值转化为该点到达最近的背景点的距离。
  欧氏距离变换在数字图像处理中的应用范围很广泛,尤其对于图像的骨架提取,是一个很好的参照。
  所谓欧氏距离变换,是指对于一张二值图像(再次我们假定白色为前景色,黑色为背景色),将前景中的像素的值转化为该点到达最近的背景点的距离。
  欧氏距离变换在数字图像处理中的应用范围很广泛,尤其对于图像的骨架提取,是一个很好的参照。
  ========
  欧氏距离:(∑(Xi-Yi)2)1/2,即两项间的差是每个变量值差的平方和再平方根,目的是计算其间的整体距离即不相似性。
  我们熟悉的 欧氏距离虽然很有用,但也有明显的缺点。它将样品的不同属性(即各指标或各变量)之间的差别等同看待,这一点有时不能满足实际要求。例如,在教育研究中, 经常遇到对人的分析和判别,个体的不同属性对于区分个体有着不同的重要性。因此,有时需要采用不同的距离函数。


3.马氏距离:
马氏距离是由印度统计学家马哈拉诺比斯(P. C. Mahalanobis)提出的,表示数据的协方差距离。它是一种有效的计算两个未知样本集的相似度的方法。与欧式距离不同的是它考虑到各种特性之间的联系(例如:一条关于身高的信息会带来一条关于体重的信息,因为两者是有关联的)并且是尺度无关的(scale-invariant),即独立于测量尺度。对于一个均值μ,为协方差矩阵为Σ的多变量向量,其马氏距离为((x-μ)'Σ^(-1)(x-μ))^(1/2)。
马氏距离也可以定义为两个服从同一分布并且其协方差矩阵为Σ的随机变量与的差异程度:
  如果协方差矩阵为单位矩阵,那么马氏距离就简化为欧式距离,如果协方差矩阵为对角阵,则其也可称为正规化的欧氏距离'.
  其中σi 是 xi 的标准差.
马氏优缺点:
  1)马氏距离的计算是建立在总体样本的基础上的,这一点可以从上述协方差矩阵的解释中可以得出,也就是说,如果拿同样的两个样本,放入两个不同的总体中,最后计算得出的两个样本间的马氏距离通常是不相同的,除非这两个总体的协方差矩阵碰巧相同;
  2)在计算马氏距离过程中,要求总体样本数大于样本的维数,否则得到的总体样本协方差矩阵逆矩阵不存在,这种情况下,用欧式距离计算即可。
  3)还有一种情况,满足了条件总体样本数大于样本的维数,但是协方差矩阵的逆矩阵仍然不存在,比如三个样本点(3,4),(5,6)和(7,8),这种情况是因为这三个样本在其所处的二维空间平面内共线。这种情况下,也采用欧式距离计算。
  4)在实际应用中“总体样本数大于样本的维数”这个条件是很容易满足的,而所有样本点出现3)中所描述的情况是很少出现的,所以在绝大多数情况下,马氏距离是可以顺利计算的,但是马氏距离的计算是不稳定的,不稳定的来源是协方差矩阵,这也是马氏距离与欧式距离的最大差异之处。
  优点:它不受量纲的影响,两点之间的马氏距离与原始数据的测量单位无关;由标准化数据和中心化数据(即原始数据与均值之差)计算出的二点之间的马氏距离相同。马氏距离还可以排除变量之间的相关性的干扰。缺点:它的缺点是夸大了变化微小的变量的作用。
  如果用dij表示第i个样品和第j个样品之间的距离,那么对一切i,j和k,dij应该满足如下四个条件:
  ①当且仅当i=j时,dij=0
  ②dij>0
  ③dij=dji(对称性)
  ④dij≤dik+dkj(三角不等式)
  显然,欧氏距离满足以上四个条件。满足以上条件的函数有多种,本节将要用到的马氏距离也是其中的一种。
  第i个样品与第j个样品的马氏距离dij用下式计算:
  dij =((x i 一x j)TS-1(x i一xj) )1/2(T、-1、1/2都是上标)
  其中,T表示转置,x i 和x j分别为第i个和第j个样品的m个指标所组成的向量,S为样本协方差矩阵。

U.S. Stock Futures Advance as Housing Construction Gains

Play
Morning Movers: Sales Misses, BOJ, Corning, Gold
U.S. stock-index futures rose, indicating the Standard & Poor’s 500 Index (SPX) will gain after a two-day drop, as data showed new housing construction rose and the Bank of Japan increased its asset-purchase target.
Sept. 19 (Bloomberg) -- Jason Trennert, chief investment strategist at Strategas Research Partners, talks about the outlook for U.S. stocks. Trennert speaks with Tom Keene and Sara Eisen on Bloomberg Television's "Surveillance." (Source: Bloomberg)
Sept. 17 (Bloomberg) -- On today's Insight & Action, Adam Johnson reviews which makes the best investment; stocks, bonds or a house. He speaks on Bloomberg Television's "Street Smart." (Source: Bloomberg)
Sept. 19 (Bloomberg) -- In today's "Single Best Chart," Scarlet Fu displays the percentage of long-term U.S. Treasures held by the Federal Reserve. She speaks on Bloomberg Television's "Bloomberg Surveillance."
Corning Inc. (GLW) gained 3.2 percent after Goldman Sachs Group Inc. advised investors to buy the stock. AutoZone Inc. (AZO) fell 3.9 percent as fourth-quarter revenue trailed analyst estimates.
Futures on the Standard & Poor’s 500 Index expiring in December rose 0.1 percent to 1,454.6 at 8:39 a.m. in New York. Dow Jones Industrial Average futures gained 17 points, or 0.1 percent, to 13,516 today.
“In terms of the growth story, unemployment and housing in the U.S. are the twin drags which need resolving,” Richard Hunter, head of equities at Hargreaves Lansdown Plc in London, wrote in an e-mail. “Japan has announced its own stimulus, with speculation now turning as to whether the Chinese will follow suit. These moves are all supportive for equities and provide investors with a welcome opportunity to think about growth rather than debt.”
The S&P 500 rallied 15 percent from its most-recent low on June 1 through Sept. 14, when it reached its highest level since December 2007. The index is trading at 14.1 times the estimated earnings of its constituent members, close to the highest price multiple since the end of 2010, data compiled by Bloomberg show.

Property Market

New-house construction rose in August, boosted by the strongest pace of single-family home starts in more than two years that shows an improving real estate market. Beginning construction climbed 2 .3 percent to a 750,000 annual rate, less than forecast and restrained by a drop in the building of apartments, from a revised 733,000 annual pace in July, Commerce Department figures showed today in Washington.
At 10 a.m., the National Association of Realtors may report sales of previously owned houses climbed 2 percent to a 4.56 million annual rate in August from 4.47 million the prior month, the survey median showed.
Japan’s central bank unexpectedly expanded its asset- purchase target by 10 trillion yen ($126 billion) to 55 trillion yen, seeking to avoid a further slowdown in the world’s third- largest economy.
Corning, the world’s largest maker of glass for flat-panel televisions, jumped 3.2 percent to $13.14 after Goldman Sachs raised its recommendation for the stock to buy from neutral.
Cracker Barrel Old Country Store Inc. (CBRL), which runs country store-themed restaurants with attached gift shops, added 7.1 percent to $68.11 after forecasting fiscal year 2013 profit of as much as $4.70 a share.
AutoZone fell 3.9 percent to $343.75. The Memphis, Tennessee-based auto-parts retailer reported fourth-quarter revenue of $2.76 billion, missing the $2.8 billion average of analyst estimates compiled by Bloomberg.
To contact the reporters on this story: Jonathan Morgan in Frankfurt at

No comments:

Post a Comment