88久久精品无码一区二区毛片 ,久久精品国产WWW456C0M,四虎影视久久久免费观看

matlab函數(shù)fminunc:求無(wú)約束優(yōu)化的最優(yōu)解

例：K-means Hashing提供的代碼optimize_one_iter函數(shù)內(nèi)部用到fminunc，代價(jià)函數(shù)是eval_one_center函數(shù)，看eval_one_center函數(shù)的注釋是this function computes the cost of Eqn. (8) given a center c_j

posted @ 2015-08-13 12:59 杰哥閱讀(877) | 評(píng)論 (0) | 編輯收藏

[zz] 機(jī)器視覺(jué)開(kāi)源代碼集合（轉(zhuǎn)載）

摘要: http://xilinx.eetrend.com/article/8919 一、特征提取Feature Extraction： SIFT [1] [Demo program][SIFT Library] [VLFeat]PCA-SIFT [2] [Project]Affine-SIFT [3] [Project]SURF [4] [OpenSURF] [Matlab Wrapper]Affi... 閱讀全文

posted @ 2015-08-13 10:19 杰哥閱讀(1196) | 評(píng)論 (0) | 編輯收藏

計(jì)算機(jī)視覺(jué)-機(jī)器學(xué)習(xí)領(lǐng)域近兩年部分綜述文章

http://www.valseonline.org/thread-505-1-1.html

【2015】

[1]. E.Sariyanidi, H. Gunes, A. Cavallaro, Automatic Analysisof Facial Affect: A Survey of Registration, Representation, and Recognition,IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 37, NO. 6,JUNE 2015

2015-1. Automatic Analysis of Facial Affect A Survey of Registration, Representa.pdf(1.51 MB, 下載次數(shù): 6)

[2]. T. Li,H. Chang, M. Wang, B.B. Ni, R.C. Hong, S.C. Yan, CrowdedScene Analysis: A Survey, IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FORVIDEO TECHNOLOGY, VOL. 25, NO. 3, MARCH 2015

2015-2. Crowded Scene Analysis A Survey.pdf(5.42 MB, 下載次數(shù): 19)

[3]. Z.Zhang, Y. Xu, J. Yang, X.L. Li, D. Zhang, A Survey ofSparse Representation: Algorithms and Applications, IEEE ACCESS, date ofpublication May 6, 2015

2015-3. A Survey of Sparse Representation.pdf(4.8 MB, 下載次數(shù): 0)

[4]. J.Galbally, S. Marcel, J. Fierrez, Biometric AntispoofingMethods: A Survey in Face Recognition, IEEE ACCESS, date of publicationDecember 18, 2014

2015-4. Biometric Antispoofing Methods A Survey in Face Recognition.pdf(10.43 MB, 下載次數(shù): 0)

[5]. B.Tian, B. T. Morris, M. Tang, Y.Q. Liu, Y. J. Yao, C. Guo, D.Y. Shen, S.H. Tang, Hierarchical and Networked Vehicle Surveillance in ITS:A Survey, IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL.16, NO. 2, APRIL 2015

2015-5. Hierarchical and Networked Vehicle Surveillance in ITS A Survey.pdf(1.12 MB, 下載次數(shù): 0)

[6]. A. Betancourt,P. Morerio, C. S. Regazzoni, and M. Rauterberg, TheEvolution of First Person Vision Methods: A Survey, IEEE TRANSACTIONS ONCIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, VOL. 25, NO. 5, MAY 2015

2015-6. The Evolution of First Person Vision Methods A Survey.pdf(2.69 MB, 下載次數(shù): 0)

[7]. L.Shao, F. Zhu, and X.L. Li, Transfer Learning for VisualCategorization: A Survey, IEEE TRANSACTIONS ON NEURAL NETWORKS ANDLEARNING SYSTEMS, VOL. 26, NO. 5, MAY 2015

2015-7. Transfer Learning for Visual Categorization A Survey.pdf(3.11 MB, 下載次數(shù): 0)

【2014】

[8]. S. Fu,H. B. He, Z.G. Hou, Learning Race from Face: A Survey, IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 36, NO.12, DECEMBER 2014

2014-1. Learning Race from Face A Survey.pdf(2.55 MB, 下載次數(shù): 15)

[9]. H.L. Zhou,A. Mian, L. Wei, D. Creighton, M. Hossny, and S. Nahavandi, Recent Advances on Singlemodal and Multimodal FaceRecognition: A Survey, IEEE TRANSACTIONS ON HUMAN-MACHINE SYSTEMS, VOL.44, NO. 6, DECEMBER 2014

2014-2. Recent Advances on Singlemodal and Multimodal Face Recognition A Survey.pdf(669.19 KB, 下載次數(shù): 0)

【2013】

[10]. O. D. Lara, M.A. Labrador, A Survey on Human Activity Recognition using WearableSensors, IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 3,THIRD QUARTER 2013

2013-1. A Survey on Human Activity Recognition using Wearable Sensors.pdf(1.65 MB, 下載次數(shù): 8)

[11]. A. Sotiras, C. Davatzikos, Nikos. Paragios, Deformable Medical Image Registration: A Survey, IEEETRANSACTIONS ON MEDICAL IMAGING, VOL. 32, NO. 7, JULY 2013

2013-2.Deformable Medical Image Registration A Survey.pdf(1.3 MB, 下載次數(shù): 11)

[12]. A. Alrahayfeh, M. Faezipour, Eye Tracking and Head Movement Detection: A State-of-ArtSurvey, IEEE Journal of Translational Engineering in Health andMedicine, 2013

2013-3. Eye Tracking and Head Movement Detection A State-of-Art Survey.pdf(2.22 MB, 下載次數(shù): 5)

[13]. P.V.K. Borges, N. Conci, and A. Cavallaro, Video-Based Human Behavior Understanding: A Survey, IEEETRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, VOL. 23, NO. 11,NOVEMBER 2013

2013-4. Video-Based Human Behavior Understanding A Survey.pdf(15.27 MB, 下載次數(shù): 46)

posted @ 2015-08-12 12:09 杰哥閱讀(255) | 評(píng)論 (0) | 編輯收藏

如何分析這兩個(gè)集合的分布的相似程度？

降維(比如PCA或者random projection) -> KDE(kernel density estimation)來(lái)估算密度 -> KL divergence

posted @ 2015-08-09 13:05 杰哥閱讀(415) | 評(píng)論 (0) | 編輯收藏

matlab加速: bsxfun

%將訓(xùn)練數(shù)據(jù)和測(cè)試數(shù)據(jù)都去中心化
X = traindata;
label = traingnd;
m = mean(X);
X_zm = bsxfun(@minus, X, m);
traindata_zm = bsxfun(@minus, traindata, m);
testdata_zm = bsxfun(@minus, testdata, m);

matlab函數(shù) bsxfun淺談（轉(zhuǎn)載）
http://blog.sina.com.cn/s/blog_9e67285801010ttn.html

網(wǎng)上關(guān)于bsxfun的東西不多，今天需要看到一個(gè)，由于原博文插入的圖片顯示不出來(lái)，于是筆者大發(fā)善心進(jìn)行了contrl+V 以及alt+ctrl+A的操作，供大家交流學(xué)習(xí)。

bsxfun是一個(gè)matlab自版本R2007a來(lái)就提供的一個(gè)函數(shù)，作用是”applies an element-by-element binary operation to arrays a and b, with singleton expansion enabled.”

舉個(gè)例子。假設(shè)我們有一列向量和一行向量。

a = randn(3,1), b = randn(1,3) a = -0.2453 -0.2766 -0.1913 b = 0.6062 0.5655 0.9057

我們可以很簡(jiǎn)單的使用matlab的外乘c=a*b來(lái)得到，如圖 bsxfun淺談（轉(zhuǎn)載）" o:button="t" target='"_blank"' o:spid="_x0000_i1025">bsxfun淺談（轉(zhuǎn)載）" src="file:///C:\Users\jie\AppData\Local\Temp\msohtmlclip1\01\clip_image001.jpg">
但如果我們想用”外加”呢？也就是說(shuō)把上式求解過(guò)程中的乘號(hào)換做加號(hào)？

這時(shí)我們可以用c=bsxfun(@plus,a,b)來(lái)實(shí)現(xiàn)。

bsxfun的執(zhí)行是這樣的，如果a和b的大小相同，那么c=a+b. 但如果有某維不同，且a或b必須有一個(gè)在這一維的維數(shù)為1, 那么bsxfun就將少的這個(gè)虛擬的復(fù)制一些來(lái)使與多的維數(shù)一樣。在我們這里，b的第一維只有1（只一行），所以bsxfun將b復(fù)制3次形成一個(gè)3×3的矩陣，同樣也將a復(fù)制成3×3的矩陣。這個(gè)等價(jià)于c=repmat(a,1,3)+repmat(b,3,1)。這里

repmat(a,1,3) ans = -0.2453 -0.2453 -0.2453 -0.2766 -0.2766 -0.2766 -0.1913 -0.1913 -0.1913

repmat是顯式的復(fù)制，當(dāng)然帶來(lái)內(nèi)存的消耗。而bsxfun是虛擬的復(fù)制，實(shí)際上通過(guò)for來(lái)實(shí)現(xiàn)，等效于for(i=1:3),for(j=1:3),c(i,j)=a(i)+b(j);end,end。但bsxfun不會(huì)有使用matlab的for所帶來(lái)額外時(shí)間。實(shí)際驗(yàn)證下這三種方式

>> c = bsxfun(@plus,a,b) c = 0.3609 0.3202 0.6604 0.3296 0.2889 0.6291 0.4149 0.3742 0.7144 >> c = repmat(a,1,3)+repmat(b,3,1) c = 0.3609 0.3202 0.6604 0.3296 0.2889 0.6291 0.4149 0.3742 0.7144 >> for(i=1:3),for(j=1:3),c(i,j)=a(i)+b(j);end,end,c c = 0.3609 0.3202 0.6604 0.3296 0.2889 0.6291 0.4149 0.3742 0.7144

從計(jì)算時(shí)間上來(lái)說(shuō)前兩種實(shí)現(xiàn)差不多，遠(yuǎn)高于for的實(shí)現(xiàn)。但如果數(shù)據(jù)很大，第二種實(shí)現(xiàn)可能會(huì)有內(nèi)存上的問(wèn)題。所以bsxfun最好。

下面看一個(gè)更為實(shí)際的情況。假設(shè)我們有數(shù)據(jù)A和B, 每行是一個(gè)樣本，每列是一個(gè)特征。我們要計(jì)算高斯核，既：

這里@plus是加法的函數(shù)數(shù)柄，相應(yīng)的有減法@minus, 乘法@times, 左右除等，具體可見(jiàn) doc bsxfun.

k(||x-xc||)=exp{- ||x-xc||^2/(2*σ)^2) } 其中xc為核函數(shù)中心,σ為函數(shù)的寬度參數(shù) , 控制了函數(shù)的徑向作用范圍。

當(dāng)然可以用雙重for實(shí)現(xiàn)（如果第一直覺(jué)是用三重for的話…）。