Face Recognition CS 663

Size: px
Start display at page:

Download "Face Recognition CS 663"

Transcription

1 Face Recognton CS 663

2 Importance of face recognton The most common way for humans to recognze each other Study of the process of face recognton has applcatons n () securty/survellance/authentcaton, () understandng of vsual psychology, (3) automated taggng on Facebook

3 Face recognton: problem statement Gven a database of face mages of people, and a new test mage, answer the followng queston: The test mage contans the face of whch ndvdual from the database?

4

5 A naïve method Compare the test mage wth each database mage n terms of SSD (sum of squared dfferences). Choose the closest match (.e., n terms of squared dfference)! I J j ( I j J j ) Ths method s fraught wth problems!

6 Challenges n face recognton: detecton Where s (are) the face(s) n the pcture?

7 Challenges n face recognton: pose varaton Images are D, face s a 3D object. There wll be out of plane rotatons (profle versus front), change n apparent sze due to change n dstance from camera.

8 Challenges n face recognton: llumnaton varaton Multple lght sources Change n drecton of lghtng sources Change n ntensty/color/type of lght source Shadows, specular reflectons

9 Challenges n face recognton: epresson varaton Vared epressons: smle, anger, frown, sadness, surprse, closed eyes, confuson, etc.

10 Challenges n face recognton: age varaton

11 Challenges n face recognton: varaton of facal accessores Spectacles, beard and moustache, scarves, ear-rngs, change of har-styles, etc.

12 Challenges n face recognton: carcatures/pantngs

13 Challenges n face recognton: blur/nose/scanner artfacts

14 And more! Even gnorng changes of pose, llumnaton, epresson, etc., we tend to look dfferent at dfferent perods of tme! Recognton stll remans a challenge!

15 Face Recognton system: block dagram () Collect database of face mages of people. Record one or multple mages per person (called gallery mage(s) of the person) (4) Label the features etracted from the mage wth that person s dentty, and store n a database () ormalze the mages: (manually) crop out the face from the overall mage background, correct for pose varaton or lghtng changes TRAIIG PHASE! (3) Etract relevant features from the normalzed face mage (more on ths later!)

16 Face Recognton system: block dagram () Collect an mage of the person whose dentty s to be determned* called the probe mage. In most cases the tme gap between acquston of probe and gallery mages s sgnfcant (months/years) () ormalze the probe mage: (manually) crop out the face from the overall mage background, correct for pose varaton or lghtng changes (3) Etract the same features from the normalzed face mage (more on ths later!) as from the gallery mages (4) Fnd the gallery mage whose features most closely match (nearest neghbor search) those of the probe mage. That tells you the dentty. TESTIG PHASE! *For now, we assume that the person whose dentty was to be determned, has mages recorded n the gallery database

17 Problems related to face recognton Face verfcaton: gven two face mages, determne whether they belong to the same ndvdual (wthout concern for the ndvdual s dentty)..

18 Problems related to face recognton Ethncty/gender dentfcaton from face mages Is the gven face mage a photo or s t a pantng/carcature?

19 What features to etract for face recognton? Many methods () Detect vsble features: eyes, nose, mouth, hgh-ponts of the cheek, chn, eyebrows, etc. ot very robust! () Statstcal holstc approaches: etract features usng statstcal method. These features may not necessarly have a physcal nterpretaton (n terms of, say, eyes, nose, mouth, etc.)

20 Egenfaces! We focus on the latter group of technques n these lectures. One such technque for face recognton called Egenfaces uses a statstcal method called Prncpal Components Analyss (PCA).

21 Prncpal Components Analyss (PCA) Consder vectors (or ponts), each contanng d elements, each represented as a column vector. d We say:, R. d could be very large lke 50,000 or more. Our am s to etract some k features from each, k << d. Effectvely we are projectng the orgnal vectors from a d-dmensonal space to a k-dmensonal space. Ths s called dmensonalty reducton. PCA s one method of dmensonalty reducton.

22 PCA How do we pck the k approprate features? We look nto a noton of compressblty how much can the data be compressed, allowng for some small errors.

23 PCA: Algorthm. Compute the mean of the gven ponts:. Deduct the mean from each pont: 3. Compute the covarance matr of these mean-deducted ponts: d d R R,, postve- semdefnte and t s a symmetrcmatr, s : :, C C ) )( ( C ote R ote d d T T

24 PCA: algorthm 4. Fnd the egenvectors of C: CV VΛ, V R dd, Λ R dd V matr of egenvecto rs (each column s an egenvecto r), Λ dagonal matr of egenvalue s ote : V s an orthonormal matr (.e. VV V V I), as C s a covarance and hence t s symmetrc. ote : Λ contans non - negatve values on the dagonal (egen - values) T T matr 5. Etract the k egenvectors correspondng to the k largest egenvalues. Ths s called the etracted egenspace: Vˆ k V(:,: k) There s an mplct assumpton here that the frst k ndces ndeed correspond to the k largest egenvalues. If that s not true, you would need to pck the approprate ndces.

25 PCA: algorthm 6. Project each pont onto the egenspace, gvng a vector of k egen-coeffcents for that pont. α As V s k Vˆ Vα Vˆ α k k T k, α k R orthonormal, V(:,) α Vˆ k (:,) α k k ; α V we have () V(:,) α () Vˆ k T, α (:,) α R ()... V(:, d) α k d ()... Vˆ k ( d) (:, d) α k ( k) We are representng each face as a lnear combnaton of the k egenvectors correspondng to the k largest egenvalues. The coeffcents of the lnear combnaton are the egen-coeffcents. ote that α k s a vector of the egencoeffcents of the -th sample pont, and t has k elements. The j-th element of ths vector s denoted as α k (j).

26 PCA and Face Recognton: Egen-faces Consder a database of cropped, frontal face mages (whch we wll assume are algned and under the same llumnaton). These are the gallery mages. We wll reshape each such mage (a D array of sze H W after croppng) to form a column vector of d = HW elements. Each mage wll be a vector, as per the notaton on the prevous two sldes. And then carry out the s steps mentoned before. The egenvectors that we get n ths case are called egenfaces. Each egenvector has d elements. If you reshape those egenvectors to form mages of sze H W, those mages look lke (fltered!) faces.

27 Eample A face database jh78_caj65/

28 Top 5 Egen-faces for ths database! jh78_caj65/

29 PCA and Face recognton: Egenfaces For each gallery mage, you compute the egen-coeffcents. You then store the egencoeffcents and the dentty of the person n a database. You also store n the database. Durng the testng phase, you are gven a probe mage (say) z p n the form of a column vector of HW elements. You deduct the mean mage from z p : z p z p Vˆ k,

30 PCA and Face recognton: Egenfaces You then project the mean-deducted face mage onto the egen-space: α p Vˆ T k z p Egen-coeffcents of the probe mage z p. ow, compare α p wth all the α k (egen-coeffcents of the gallery mages) n the database. Fnd the closest match n terms of the squared dstance between the egen-coeffcents. That gves you the dentty (see net slde).

31 PCA and Face recognton: Egenfaces jp arg mn l α p α l Egen-coeffcents of the probe mage z p. Egen-coeffcents of the l-th gallery mage l. ote: other dstance measures (dfferent from sum of squared dfferences) may also be employed. One eample s sum of absolute dfferences, gven as follows: α p α l Another could be normalzed dot product (and ths dstance measure should be mamzed!): α p αl α p α l

32 PCA and Face recognton: egenfaces The egen-face mages contan more and more hgh frequency nformaton as the correspondng egen-values decrease. Although PCA s a technque known for a long tme, t s applcaton n face recognton was poneered by Turk and Pentland n a classc paper n 99. M. Turk and A. Pentland (99). Egenfaces for recognton, Journal of Cogntve euroscence, 3(): 7 86.

33

34 PCA and Face recognton: egenfaces We can regard the k egenfaces as key sgnatures. We epress each face mage as a lnear combnaton of these egenfaces,.e. the average face + (say) 3 tmes egenface + (say) 5 tmes egenface + (say) - tmes egenface 3 and so on. (note: 3,5,- here are the egencoeffcents, and some of them can be negatve). Vα Vˆ α k k V(:,) α Vˆ k (:,) α k () () V(:,) α Vˆ k (:,) α ()... k ()... V(:, d) α Vˆ k ( d) (:, d) α k ( k)

35 One word of cauton: Egen-faces The algorthm descrbed earler s computatonally nfeasble for egen-faces, as t requres storage of a d d Covarance matr (d the number of mage pels - could be more than 0,000). And the computaton of the egenvectors of such a matr s a O(d 3 ) operaton! We wll study a modfcaton to ths that wll brng down the computatonal cost drastcally.

36 Egen-faces: reducng computatonal complety. Consder the covarance matr: C ote : C s a symmetrcmatr, T ( )( and t s ), ote : C R postve- semdefn te It wll requre too much memory f d s large, and computng ts egenvectors wll be a horrendous task! Consder the case when s much less than d. Ths s very common n face recognton applcatons. The number of tranng mages s usually much smaller than the sze of the mage. T dd

37 Egen-faces: reducng computatonal complety. In such a case, the rank of C s at the most -. So C wll have at the most - non-zero egenvalues. We can wrte C n the followng way: C X [ T... XX ] R T d, where

38 Back to Egen-faces: reducng computatonal complety. Consder the matr X T X (sze ) nstead of XX T (sze d d). Its egenvectors are of the form: X T Xw w, w R XX T ( Xw) ( Xw)[pre multplyng by X] Xw s an egenvector of C=XX T! Computng all egenvectors of C wll now have a complety of only O( 3 ) for computaton of the egenvectors of X T X + O( d) for computaton of Xw from each w = total of O( 3 + d ) whch s much less than O(d 3 ). ote that C has at most only mn(-,d) egenvectors correspondng to non-zero egen-values (why?).

39 Egenfaces: Algorthm ( << d case). Compute the mean of the gven ponts:. Deduct the mean from each pont: 3. Compute the followng matr: postve- semdefn te and t s s a symmetrcmatr, : ] [,, L... X L X X L ote R R d T d d R R,,

40 Egen-faces: Algorthm ( << d case) 4. Fnd the egenvectors of L: LW WΓ, W egenvecto rs, Γ egenvalues, T WW I 5. Obtan the egenvectors of C from those of L: d V XW, XR, WR, VR 6. Unt-normalze the columns of V. 7. C wll have at most only egenvectors correspondng to non-zero egen-values*. Out of these you pck the top k (k < ) correspondng to the largest egen-values. * Actually ths number s at most - ths s due to the mean subtracton, else t would have been at most.

41 Eample A face database jh78_caj65/

42 Top 5 Egen-faces for ths database! jh78_caj65/

43 Eample The Yale Face database

44 Top 5 egenfaces from the prevous database Reconstructon of a face mage usng the top,8,6,3,,04 egenfaces (.e. k vared from to 04 n steps of 8) Vα ˆ k l Vˆ (:, l) α k ( l)

45 What f both and d are large? Ths can happen, for eample, f you wanted to buld an egenspace for face mages of all people n Mumba. Dvde people nto coherent groups based on some vsual attrbutes (eg: gender, age group etc) and buld separate egenspaces for each group.

46 PCA: A closer look PCA has many applcatons apart from face/object recognton n mage processng/computer vson, statstcs, econometrcs, fnance, agrculture, and you name t! Why PCA? What s specal about PCA? See the net sldes!

47 PCA: what does t do? It fnds k perpendcular drectons (all passng through the mean vector) such that the orgnal data are appromated as accurately as possble when projected onto these k drectons. We wll see soon why these k drectons are egenvectors of the covarance matr of the data!

48 PCA Look at ths scatter-plot of ponts n D. The ponts are hghly spread out n the drecton of the lght blue lne.

49 PCA Ths s how the data would look f they were rotated n such a way that the major as of the ellpse (the lght blue lne) now concded wth the Y as. As the spread of the X coordnates s now relatvely nsgnfcant (observe the aes!), we can appromate the rotated data ponts by ther projectons onto the Y-as (.e. ther Y coordnates alone!). Ths was not possble pror to rotaton!

50 PCA As we could gnore the X-coordnates of the ponts post rotaton and represent them just by the Y-coordnates, we have performed some sort of lossy data compresson or dmensonalty reducton. The job of PCA s to perform such a rotaton as shown on the prevous two sldes!

51 PCA Am of PCA: Fnd the lne passng through the sample mean (.e. ), such that the projecton of any mean-deducted pont onto e, most accurately appromates t. Projecton of - ontoe s T a e( - ) e a e, a R, Error of appromaton (ae)( ote: Here e s a unt vector. ) a e

52 PCA Summng up over all ponts, we get: a J ) ( ) ( ) ( appromaton totalerror of Sum e e ) ( e T a a ) ( ) ( e e T a a ) ( ) ( e e T a a T a a a a )) (,( e

53 PCA a J ) ( appromaton totalerror of Sum e Ths term s proportonal to the varance of the data ponts when projected onto the drecton e. t t ) )( ( e e t )) ( ( e ) ) ( (where C S Se e t

54 J ( e) t e Se PCA Mnmzng J( e) w.r.t. e s equvalent to mamzng Independent of the drecton e t e Se w.r.t. e. We use the method of Lagrange multplers todo so, whle smultaneously t mposng the constrant that e e. See append for detals So we have to take the dervatve of the followng modfed functon w.r.t. e (and set t to0) ~ t t J ( e) e Se ( e e -) ~ Takng dervatve of J ( e) w.r.t. e and settng t to0, we get Se e, soe s an egen - vector of S. t t Ase Se and we wsh to mamze e Se, wechoose e to be the egen - vector correspondng to the mamum egenvalue of S.

55 PCA PCA thus projects the data onto that drecton that mnmzes the total squared dfference between the data-ponts and ther respectve projectons along that drecton. Ths equvalently yelds the drecton along whch the spread (or varance) wll be mamum. Why? ote that the egenvalue of a covarance matr tells you the varance of the data when projected along that partcular egenvector: Se e t e Se t e Se t ( e ( )) Ths term s proportonal to the varance of the data when projected along e.

56 PCA But for most applcatons (ncludng face recognton), just a sngle drecton s absolutely nsuffcent! We wll need to project the data (from the hghdmensonal,.e. d-dmensonal space) onto k (k << d) dfferent mutually perpendcular drectons. What s the crteron for dervng these drectons? We seek those k drectons for whch the total reconstructon error of all the mages when projected on those drectons s mnmzed.

57 PCA We seek those k drectons for whch the total reconstructon error of all the mages when projected on those drectons s mnmzed. k t J ({ ej} j ) ( ) (ej( )) e j k j One can prove that these k drectons wll be the egenvectors of the S matr (equvalently covarance matr of the data) correspondng to the k-largest egenvalues. These k drectons form the egen-space. If the egenvalues of S are dstnct, these k drectons are defned unquely (up to a sgn factor)

58 PCA One can prove that these k drectons wll be the egenvectors of the S matr (equvalently covarance matr of the data) correspondng to the k-largest egenvalues. These k drectons form the egen-space. Sketch of the proof: Assume we have found e and are lookng for e (where e s perpendcular to e and e has unt magntude). Wrte out the objectve functon wth the two constrants. Mnmze t and do some algebra to see that e s the egenvector of S wth the second largest egenvalue. Proceed smlarly for other drectons.

59 How to pck k n an actual applcaton? Tral and error. Usually between 50 to 00 for mages of sze Dvde your tranng set nto two parts A and B (B s usually called the valdaton set). Pck the value of k that gves the best recognton rate on B when you tran on A. Stck to that value of k. ote: a larger k mples a better reconstructon but t may even cause a decrease n the recognton accuracy!

60 How to pck k n an actual applcaton? ote: a larger k mples a better reconstructon but t may even cause a decrease n the recognton accuracy! Why because throwng out some of the egenvectors may lead to flterng of the data, removng some unnecessary artfacts for eample.

61 Some observatons about PCA for face mages The matr V (wth all columns) s orthonormal. Hence the squared error between any mage and ts appromaton usng just top k egenvectors s gven by: ~ α α ~ d α jk V ( α ( why?) ( why?) α ~ ) ( why?) Ths error s small on an average for a well-algned group of face mages we wll see why on the net slde.

62 The egenvalues of the covarance matr typcally decay fast n value (f the faces were properly normalzed). ote that the j-th egenvalue s proportonal to the varance of the j-th egencoeffcent,.e. t e Se j j j e ( t j )( t ) e j ( ) E( ) What ths means s that the data have low varance when projected along most of the egenvectors,.e. effectvely the data are concentrated n a lower-dmensonal subspace of the d-dmensonal space. j

63 Person-specfc egen-faces So far, we bult one egen-space for the whole database consstng of multple mages each of multple people. Alternatve approach: Construct one egen-space (.e. a set of some k egenvectors) for each of the M people. Assume we have multple gallery mages per person, possbly under dfferent poses, llumnaton condtons and epressons.

64 Person-specfc egen-space Let the egen-space for the r-th person be denoted as V (r) k and the correspondng mean (r) mage be denoted as. The egen-coeffcents of a probe mage onto the r-th egen-space are gven as: β (r) p V T (r) k (z p (r) ) For every r, determne the followng: d( r) z p r Vˆ (r) (r) k βp A measure of how well k egen-vectors from the egen-space for person r, are capable of reconstructng the probe mage. ote that k number of gallery mages for the r-th person (why?)

65 Person-specfc egen-space An alternatve dstance measure s: d( r) β (r) (r) β p Average egen-coeffcent vector of all gallery mages belongng to the r-th person The dentty of the probe mage s gven by the egen-space r for whch d(r) was the mnmum.

66 Face recognton under vared lghtng The appearance varaton over mages of the same person under dfferent llumnatons s much greater than the varaton over mages of dfferent people under the same lghtng condton!

67 Face recognton under vared lghtng Gven a database of mages of M people each under L lghtng condtons, create an egenspace and remove the top 3 prncpal components for better face recognton! Why we wll understand the real reason n a computer vson class!

68 A word of clarfcaton PCA magc. It has no n-bult ablty to perform pose normalzaton. For pose normalzaton, you can buld pose-specfc egenspaces or person-specfc egen-spaces usng multple mages of a person under dfferent poses. For (partal) llumnaton normalzaton, you can use the trck mentoned on the prevous slde. However, n general pose normalzaton s a major problem n face recognton from D mages.

69 Another word of clarfcaton PCA at ts core s a reconstructon algorthm. It s not a classfcaton algorthm and s not desgned to be one. But t showed very good results for face recognton and hence became popular n ths communty. There are other methods (eg: Lnear Dscrmnant Analyss) whch are desgned for the purpose of good classfcaton on face mages or other datasets.

70 PCA: Compresson of a set of mages Consder a database of mages that are smlar (eg: all are face mages, all are car mages, etc.) Buld an egen-space from some subset of these mages (could be all mages, as well) We know that these mages can often be reconstructed very well (.e. wth low error) usng just a few egenvectors.

71 PCA: Compresson of a set of mages Use ths fact for mage compresson. Orgnal data storage = d pels mages = d bytes (assume one byte per pel ntensty) = 8d bts. After PCA: k number of bts to store each egencoeffcent = 3k bts (remember k << d, eample: d ~ 50,000 and k ~ 00). Plus storage of egenvectors = 3dk bts (remember k << M as well). Plus mean mage = 8d bts. Total: 3(+d)k + 8d bts

72 PCA: Compresson of a set of mages Eample: = 5000, d = 50000, k = 00 Orgnal sze/(sze after PCA compresson) ~.. ote: we allocated 3 bts for every element of the egen-vector. Ths s actually very conservatve and you can have further savngs usng several trcks.

73 PCA: Compresson of a set of mages Ths dffers a lot from JPEG compresson. JPEG uses dscrete cosne transform (DCT) as the bass. PCA tunes the bass to the underlyng tranng data! If you change the tranng data, the bass changes! The performance of the PCA compresson algorthm wll depend on how compactly the egen-space can represent the data. We wll study mage compresson usng JPEG and other standards later on n the course.

74 Face Recognton: Other types of mages From range data called as 3D face recognton ( )

75 Face Recognton: Other types of From vdeo! mages

76 Face recognton: from the FBI

77 Concluson We studed: Face recognton and related problem statements Egenfaces algorthm faster and slower versons Dervaton of the algorthm Modfcatons for face detecton/verfcaton/person-specfc egenfaces Applcaton for compresson of mages

78 References Secton 3.8 of Pattern Classfcaton by Duda and Hart M. Turk and A. Pentland (99). "Egenfaces for recognton". Journal of Cogntve euroscence

79 Append: Covarance matr In probablty and statstcs, the epected value of a scalar random varable s called ts mean: E( ) Probablty densty functon of p( ) d The varance of a scalar random varable s the epected value of ts squared devaton around the mean: E(( ) ) ( ) p( ) d ( ) Sample values of the random varable

80 Append: Covarance matr μ k The epected value of a vector random varable s called ts mean vector: E( ), ( (), k, k (),..., d ( d)), μ k-th sample value of ( (), (),..., ( d))

81 Append: Covarance matr The varance s now replaced by a covarance matr. Each entry contans the covarance between two elements of the vector: vector element of ) ( :, )), ( ) ( ))( ( ) ( ( ))) ( ) ( ))( ( ) ( (( ) )) ( ) ( ((... ))) ( ) ( ())( () (( ) ()) () (( ())) () ())( () (( ())) () ))( ( ) ( ((.. ())) () ())( () (( ) ()) () (( th k k ote d l d k l l k k l l k k E C R d d E d d E E E d d E E E kl d d C

82 Append: A word about Lagrange Multplers Consder you want to fnd the mnmum or mamum of f(,y). ormally, you take the dervatve and set t to 0 and check the sgn of the second dervatve. ow you want to fnd the optmum subject to a constrant that h(,y) = 0.

83 Append: A word about Lagrange Multplers Physcal analogy: f you drop a pebble nto a parabola-shaped bowl, the pebble wll fall to the bottom-most pont. But f you placed a wooden plank nto the bowl, the pebble wll settle somewhere on the plank!

84 c * c c c3 c4 c5 ange-multpler.pdf Suppose we are walkng from (,y ) to (,y ) along the constrant curve h(,y) = 0. Intally, the tangent vector along h(,y) = 0 has a component along grad(f), and hence there s a decrease n the value of f(,y) as we move along h(,y) = 0. If the pont (*,y*) s a local mnmum of f(,y), a small moton along h(,y) = 0 from (*,y*) wll have no component along grad(f), else t would cause an ncrease n the value of f. Hence the tangent to h(,y) = 0 at (*,y*) wll be perpendcular to grad(f). As we move further, moton along h(,y) = 0 wll have a component along +grad(f) leadng to an ncrease n the value of f(,y). At the mnmum pont, we have grad(f) perpendcular to the tangent,.e. grad(f) and grad(h) are collnear. Hence, there ests some value λ (the Lagrange multpler ) such that we have: f ( *, y*) h( *, y*) ( f ( *, y*) h( *, y*)) 0

85 If (',y') s the etremum of small perturbaton to(',y') wll lead to no change of f - n Hence we have f( ', y' ) But here we are constraned to move only along the curve h(,y) Let s be the tangent to ths curve at a pont (', y'). The magntude f( ', y' ) s. If (', y') s an etremum of f( ', y' ) s the lmt when of 0. the change functon the magntude 0. along But thenormal to thecurve s h( ', y' ) and t s perpendcular to s. Hence we have f( ', y' ) s collnear ths curve, f( ', y' ) h( ', y' ) for some value. f n f f(,y), then a of n the perturbaton s 0. due to nfntesmal wth h( ', y' ). the value movement along then we must have 0. thecurve s gven by In our dervaton for PCA, nstead of t we have e e 0. otethat e s a vector n d t f(,y) we have e Se, and nstead of dmensonal space. h(, y) 0,

Which Separator? Spring 1

Which Separator? Spring 1 Whch Separator? 6.034 - Sprng 1 Whch Separator? Mamze the margn to closest ponts 6.034 - Sprng Whch Separator? Mamze the margn to closest ponts 6.034 - Sprng 3 Margn of a pont " # y (w $ + b) proportonal

More information

Regularized Discriminant Analysis for Face Recognition

Regularized Discriminant Analysis for Face Recognition 1 Regularzed Dscrmnant Analyss for Face Recognton Itz Pma, Mayer Aladem Department of Electrcal and Computer Engneerng, Ben-Guron Unversty of the Negev P.O.Box 653, Beer-Sheva, 845, Israel. Abstract Ths

More information

Statistical pattern recognition

Statistical pattern recognition Statstcal pattern recognton Bayes theorem Problem: decdng f a patent has a partcular condton based on a partcular test However, the test s mperfect Someone wth the condton may go undetected (false negatve

More information

Unified Subspace Analysis for Face Recognition

Unified Subspace Analysis for Face Recognition Unfed Subspace Analyss for Face Recognton Xaogang Wang and Xaoou Tang Department of Informaton Engneerng The Chnese Unversty of Hong Kong Shatn, Hong Kong {xgwang, xtang}@e.cuhk.edu.hk Abstract PCA, LDA

More information

Lecture 10 Support Vector Machines II

Lecture 10 Support Vector Machines II Lecture 10 Support Vector Machnes II 22 February 2016 Taylor B. Arnold Yale Statstcs STAT 365/665 1/28 Notes: Problem 3 s posted and due ths upcomng Frday There was an early bug n the fake-test data; fxed

More information

PHYS 705: Classical Mechanics. Calculus of Variations II

PHYS 705: Classical Mechanics. Calculus of Variations II 1 PHYS 705: Classcal Mechancs Calculus of Varatons II 2 Calculus of Varatons: Generalzaton (no constrant yet) Suppose now that F depends on several dependent varables : We need to fnd such that has a statonary

More information

A Tutorial on Data Reduction. Linear Discriminant Analysis (LDA) Shireen Elhabian and Aly A. Farag. University of Louisville, CVIP Lab September 2009

A Tutorial on Data Reduction. Linear Discriminant Analysis (LDA) Shireen Elhabian and Aly A. Farag. University of Louisville, CVIP Lab September 2009 A utoral on Data Reducton Lnear Dscrmnant Analss (LDA) hreen Elhaban and Al A Farag Unverst of Lousvlle, CVIP Lab eptember 009 Outlne LDA objectve Recall PCA No LDA LDA o Classes Counter eample LDA C Classes

More information

Lecture 3: Probability Distributions

Lecture 3: Probability Distributions Lecture 3: Probablty Dstrbutons Random Varables Let us begn by defnng a sample space as a set of outcomes from an experment. We denote ths by S. A random varable s a functon whch maps outcomes nto the

More information

CIS526: Machine Learning Lecture 3 (Sept 16, 2003) Linear Regression. Preparation help: Xiaoying Huang. x 1 θ 1 output... θ M x M

CIS526: Machine Learning Lecture 3 (Sept 16, 2003) Linear Regression. Preparation help: Xiaoying Huang. x 1 θ 1 output... θ M x M CIS56: achne Learnng Lecture 3 (Sept 6, 003) Preparaton help: Xaoyng Huang Lnear Regresson Lnear regresson can be represented by a functonal form: f(; θ) = θ 0 0 +θ + + θ = θ = 0 ote: 0 s a dummy attrbute

More information

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X Statstcs 1: Probablty Theory II 37 3 EPECTATION OF SEVERAL RANDOM VARIABLES As n Probablty Theory I, the nterest n most stuatons les not on the actual dstrbuton of a random vector, but rather on a number

More information

Generalized Linear Methods

Generalized Linear Methods Generalzed Lnear Methods 1 Introducton In the Ensemble Methods the general dea s that usng a combnaton of several weak learner one could make a better learner. More formally, assume that we have a set

More information

Kernel Methods and SVMs Extension

Kernel Methods and SVMs Extension Kernel Methods and SVMs Extenson The purpose of ths document s to revew materal covered n Machne Learnng 1 Supervsed Learnng regardng support vector machnes (SVMs). Ths document also provdes a general

More information

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography CSc 6974 and ECSE 6966 Math. Tech. for Vson, Graphcs and Robotcs Lecture 21, Aprl 17, 2006 Estmatng A Plane Homography Overvew We contnue wth a dscusson of the major ssues, usng estmaton of plane projectve

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 30 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 2 Remedes for multcollnearty Varous technques have

More information

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix Lectures - Week 4 Matrx norms, Condtonng, Vector Spaces, Lnear Independence, Spannng sets and Bass, Null space and Range of a Matrx Matrx Norms Now we turn to assocatng a number to each matrx. We could

More information

Salmon: Lectures on partial differential equations. Consider the general linear, second-order PDE in the form. ,x 2

Salmon: Lectures on partial differential equations. Consider the general linear, second-order PDE in the form. ,x 2 Salmon: Lectures on partal dfferental equatons 5. Classfcaton of second-order equatons There are general methods for classfyng hgher-order partal dfferental equatons. One s very general (applyng even to

More information

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity Week3, Chapter 4 Moton n Two Dmensons Lecture Quz A partcle confned to moton along the x axs moves wth constant acceleraton from x =.0 m to x = 8.0 m durng a 1-s tme nterval. The velocty of the partcle

More information

Chapter 11: Simple Linear Regression and Correlation

Chapter 11: Simple Linear Regression and Correlation Chapter 11: Smple Lnear Regresson and Correlaton 11-1 Emprcal Models 11-2 Smple Lnear Regresson 11-3 Propertes of the Least Squares Estmators 11-4 Hypothess Test n Smple Lnear Regresson 11-4.1 Use of t-tests

More information

= = = (a) Use the MATLAB command rref to solve the system. (b) Let A be the coefficient matrix and B be the right-hand side of the system.

= = = (a) Use the MATLAB command rref to solve the system. (b) Let A be the coefficient matrix and B be the right-hand side of the system. Chapter Matlab Exercses Chapter Matlab Exercses. Consder the lnear system of Example n Secton.. x x x y z y y z (a) Use the MATLAB command rref to solve the system. (b) Let A be the coeffcent matrx and

More information

Singular Value Decomposition: Theory and Applications

Singular Value Decomposition: Theory and Applications Sngular Value Decomposton: Theory and Applcatons Danel Khashab Sprng 2015 Last Update: March 2, 2015 1 Introducton A = UDV where columns of U and V are orthonormal and matrx D s dagonal wth postve real

More information

INF 5860 Machine learning for image classification. Lecture 3 : Image classification and regression part II Anne Solberg January 31, 2018

INF 5860 Machine learning for image classification. Lecture 3 : Image classification and regression part II Anne Solberg January 31, 2018 INF 5860 Machne learnng for mage classfcaton Lecture 3 : Image classfcaton and regresson part II Anne Solberg January 3, 08 Today s topcs Multclass logstc regresson and softma Regularzaton Image classfcaton

More information

CS 523: Computer Graphics, Spring Shape Modeling. PCA Applications + SVD. Andrew Nealen, Rutgers, /15/2011 1

CS 523: Computer Graphics, Spring Shape Modeling. PCA Applications + SVD. Andrew Nealen, Rutgers, /15/2011 1 CS 523: Computer Graphcs, Sprng 20 Shape Modelng PCA Applcatons + SVD Andrew Nealen, utgers, 20 2/5/20 emnder: PCA Fnd prncpal components of data ponts Orthogonal drectons that are domnant n the data (have

More information

Efficient, General Point Cloud Registration with Kernel Feature Maps

Efficient, General Point Cloud Registration with Kernel Feature Maps Effcent, General Pont Cloud Regstraton wth Kernel Feature Maps Hanchen Xong, Sandor Szedmak, Justus Pater Insttute of Computer Scence Unversty of Innsbruck 30 May 2013 Hanchen Xong (Un.Innsbruck) 3D Regstraton

More information

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg prnceton unv. F 17 cos 521: Advanced Algorthm Desgn Lecture 7: LP Dualty Lecturer: Matt Wenberg Scrbe: LP Dualty s an extremely useful tool for analyzng structural propertes of lnear programs. Whle there

More information

Feature Selection: Part 1

Feature Selection: Part 1 CSE 546: Machne Learnng Lecture 5 Feature Selecton: Part 1 Instructor: Sham Kakade 1 Regresson n the hgh dmensonal settng How do we learn when the number of features d s greater than the sample sze n?

More information

VQ widely used in coding speech, image, and video

VQ widely used in coding speech, image, and video at Scalar quantzers are specal cases of vector quantzers (VQ): they are constraned to look at one sample at a tme (memoryless) VQ does not have such constrant better RD perfomance expected Source codng

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

More information

Lecture 10: Dimensionality reduction

Lecture 10: Dimensionality reduction Lecture : Dmensonalt reducton g The curse of dmensonalt g Feature etracton s. feature selecton g Prncpal Components Analss g Lnear Dscrmnant Analss Intellgent Sensor Sstems Rcardo Guterrez-Osuna Wrght

More information

Notes on Frequency Estimation in Data Streams

Notes on Frequency Estimation in Data Streams Notes on Frequency Estmaton n Data Streams In (one of) the data streamng model(s), the data s a sequence of arrvals a 1, a 2,..., a m of the form a j = (, v) where s the dentty of the tem and belongs to

More information

Composite Hypotheses testing

Composite Hypotheses testing Composte ypotheses testng In many hypothess testng problems there are many possble dstrbutons that can occur under each of the hypotheses. The output of the source s a set of parameters (ponts n a parameter

More information

Linear Approximation with Regularization and Moving Least Squares

Linear Approximation with Regularization and Moving Least Squares Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...

More information

MACHINE APPLIED MACHINE LEARNING LEARNING. Gaussian Mixture Regression

MACHINE APPLIED MACHINE LEARNING LEARNING. Gaussian Mixture Regression 11 MACHINE APPLIED MACHINE LEARNING LEARNING MACHINE LEARNING Gaussan Mture Regresson 22 MACHINE APPLIED MACHINE LEARNING LEARNING Bref summary of last week s lecture 33 MACHINE APPLIED MACHINE LEARNING

More information

Chapter 8 Indicator Variables

Chapter 8 Indicator Variables Chapter 8 Indcator Varables In general, e explanatory varables n any regresson analyss are assumed to be quanttatve n nature. For example, e varables lke temperature, dstance, age etc. are quanttatve n

More information

Lecture 12: Classification

Lecture 12: Classification Lecture : Classfcaton g Dscrmnant functons g The optmal Bayes classfer g Quadratc classfers g Eucldean and Mahalanobs metrcs g K Nearest Neghbor Classfers Intellgent Sensor Systems Rcardo Guterrez-Osuna

More information

EPR Paradox and the Physical Meaning of an Experiment in Quantum Mechanics. Vesselin C. Noninski

EPR Paradox and the Physical Meaning of an Experiment in Quantum Mechanics. Vesselin C. Noninski EPR Paradox and the Physcal Meanng of an Experment n Quantum Mechancs Vesseln C Nonnsk vesselnnonnsk@verzonnet Abstract It s shown that there s one purely determnstc outcome when measurement s made on

More information

Lecture 3 Stat102, Spring 2007

Lecture 3 Stat102, Spring 2007 Lecture 3 Stat0, Sprng 007 Chapter 3. 3.: Introducton to regresson analyss Lnear regresson as a descrptve technque The least-squares equatons Chapter 3.3 Samplng dstrbuton of b 0, b. Contnued n net lecture

More information

Maximal Margin Classifier

Maximal Margin Classifier CS81B/Stat41B: Advanced Topcs n Learnng & Decson Makng Mamal Margn Classfer Lecturer: Mchael Jordan Scrbes: Jana van Greunen Corrected verson - /1/004 1 References/Recommended Readng 1.1 Webstes www.kernel-machnes.org

More information

U.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017

U.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017 U.C. Berkeley CS94: Beyond Worst-Case Analyss Handout 4s Luca Trevsan September 5, 07 Summary of Lecture 4 In whch we ntroduce semdefnte programmng and apply t to Max Cut. Semdefnte Programmng Recall that

More information

MIMA Group. Chapter 2 Bayesian Decision Theory. School of Computer Science and Technology, Shandong University. Xin-Shun SDU

MIMA Group. Chapter 2 Bayesian Decision Theory. School of Computer Science and Technology, Shandong University. Xin-Shun SDU Group M D L M Chapter Bayesan Decson heory Xn-Shun Xu @ SDU School of Computer Scence and echnology, Shandong Unversty Bayesan Decson heory Bayesan decson theory s a statstcal approach to data mnng/pattern

More information

10-701/ Machine Learning, Fall 2005 Homework 3

10-701/ Machine Learning, Fall 2005 Homework 3 10-701/15-781 Machne Learnng, Fall 2005 Homework 3 Out: 10/20/05 Due: begnnng of the class 11/01/05 Instructons Contact questons-10701@autonlaborg for queston Problem 1 Regresson and Cross-valdaton [40

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 31 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 6. Rdge regresson The OLSE s the best lnear unbased

More information

( ) [ ( k) ( k) ( x) ( ) ( ) ( ) [ ] ξ [ ] [ ] [ ] ( )( ) i ( ) ( )( ) 2! ( ) = ( ) 3 Interpolation. Polynomial Approximation.

( ) [ ( k) ( k) ( x) ( ) ( ) ( ) [ ] ξ [ ] [ ] [ ] ( )( ) i ( ) ( )( ) 2! ( ) = ( ) 3 Interpolation. Polynomial Approximation. 3 Interpolaton {( y } Gven:,,,,,, [ ] Fnd: y for some Mn, Ma Polynomal Appromaton Theorem (Weerstrass Appromaton Theorem --- estence ε [ ab] f( P( , then there ests a polynomal

More information

Week 11: Chapter 11. The Vector Product. The Vector Product Defined. The Vector Product and Torque. More About the Vector Product

Week 11: Chapter 11. The Vector Product. The Vector Product Defined. The Vector Product and Torque. More About the Vector Product The Vector Product Week 11: Chapter 11 Angular Momentum There are nstances where the product of two vectors s another vector Earler we saw where the product of two vectors was a scalar Ths was called the

More information

Answers Problem Set 2 Chem 314A Williamsen Spring 2000

Answers Problem Set 2 Chem 314A Williamsen Spring 2000 Answers Problem Set Chem 314A Wllamsen Sprng 000 1) Gve me the followng crtcal values from the statstcal tables. a) z-statstc,-sded test, 99.7% confdence lmt ±3 b) t-statstc (Case I), 1-sded test, 95%

More information

Pattern Classification

Pattern Classification Pattern Classfcaton All materals n these sldes ere taken from Pattern Classfcaton (nd ed) by R. O. Duda, P. E. Hart and D. G. Stork, John Wley & Sons, 000 th the permsson of the authors and the publsher

More information

BACKGROUND SUBTRACTION WITH EIGEN BACKGROUND METHODS USING MATLAB

BACKGROUND SUBTRACTION WITH EIGEN BACKGROUND METHODS USING MATLAB BACKGROUND SUBTRACTION WITH EIGEN BACKGROUND METHODS USING MATLAB 1 Ilmyat Sar 2 Nola Marna 1 Pusat Stud Komputas Matematka, Unverstas Gunadarma e-mal: lmyat@staff.gunadarma.ac.d 2 Pusat Stud Komputas

More information

/ n ) are compared. The logic is: if the two

/ n ) are compared. The logic is: if the two STAT C141, Sprng 2005 Lecture 13 Two sample tests One sample tests: examples of goodness of ft tests, where we are testng whether our data supports predctons. Two sample tests: called as tests of ndependence

More information

LECTURE 9 CANONICAL CORRELATION ANALYSIS

LECTURE 9 CANONICAL CORRELATION ANALYSIS LECURE 9 CANONICAL CORRELAION ANALYSIS Introducton he concept of canoncal correlaton arses when we want to quantfy the assocatons between two sets of varables. For example, suppose that the frst set of

More information

The Geometry of Logit and Probit

The Geometry of Logit and Probit The Geometry of Logt and Probt Ths short note s meant as a supplement to Chapters and 3 of Spatal Models of Parlamentary Votng and the notaton and reference to fgures n the text below s to those two chapters.

More information

Why Bayesian? 3. Bayes and Normal Models. State of nature: class. Decision rule. Rev. Thomas Bayes ( ) Bayes Theorem (yes, the famous one)

Why Bayesian? 3. Bayes and Normal Models. State of nature: class. Decision rule. Rev. Thomas Bayes ( ) Bayes Theorem (yes, the famous one) Why Bayesan? 3. Bayes and Normal Models Alex M. Martnez alex@ece.osu.edu Handouts Handoutsfor forece ECE874 874Sp Sp007 If all our research (n PR was to dsappear and you could only save one theory, whch

More information

ADVANCED MACHINE LEARNING ADVANCED MACHINE LEARNING

ADVANCED MACHINE LEARNING ADVANCED MACHINE LEARNING 1 ADVANCED ACHINE LEARNING ADVANCED ACHINE LEARNING Non-lnear regresson technques 2 ADVANCED ACHINE LEARNING Regresson: Prncple N ap N-dm. nput x to a contnuous output y. Learn a functon of the type: N

More information

Note on EM-training of IBM-model 1

Note on EM-training of IBM-model 1 Note on EM-tranng of IBM-model INF58 Language Technologcal Applcatons, Fall The sldes on ths subject (nf58 6.pdf) ncludng the example seem nsuffcent to gve a good grasp of what s gong on. Hence here are

More information

Subspace Learning Based on Tensor Analysis. by Deng Cai, Xiaofei He, and Jiawei Han

Subspace Learning Based on Tensor Analysis. by Deng Cai, Xiaofei He, and Jiawei Han Report No. UIUCDCS-R-2005-2572 UILU-ENG-2005-1767 Subspace Learnng Based on Tensor Analyss by Deng Ca, Xaofe He, and Jawe Han May 2005 Subspace Learnng Based on Tensor Analyss Deng Ca Xaofe He Jawe Han

More information

CSE 252C: Computer Vision III

CSE 252C: Computer Vision III CSE 252C: Computer Vson III Lecturer: Serge Belonge Scrbe: Catherne Wah LECTURE 15 Kernel Machnes 15.1. Kernels We wll study two methods based on a specal knd of functon k(x, y) called a kernel: Kernel

More information

Support Vector Machines. Vibhav Gogate The University of Texas at dallas

Support Vector Machines. Vibhav Gogate The University of Texas at dallas Support Vector Machnes Vbhav Gogate he Unversty of exas at dallas What We have Learned So Far? 1. Decson rees. Naïve Bayes 3. Lnear Regresson 4. Logstc Regresson 5. Perceptron 6. Neural networks 7. K-Nearest

More information

Some Comments on Accelerating Convergence of Iterative Sequences Using Direct Inversion of the Iterative Subspace (DIIS)

Some Comments on Accelerating Convergence of Iterative Sequences Using Direct Inversion of the Iterative Subspace (DIIS) Some Comments on Acceleratng Convergence of Iteratve Sequences Usng Drect Inverson of the Iteratve Subspace (DIIS) C. Davd Sherrll School of Chemstry and Bochemstry Georga Insttute of Technology May 1998

More information

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results.

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results. Neural Networks : Dervaton compled by Alvn Wan from Professor Jtendra Malk s lecture Ths type of computaton s called deep learnng and s the most popular method for many problems, such as computer vson

More information

Error Bars in both X and Y

Error Bars in both X and Y Error Bars n both X and Y Wrong ways to ft a lne : 1. y(x) a x +b (σ x 0). x(y) c y + d (σ y 0) 3. splt dfference between 1 and. Example: Prmordal He abundance: Extrapolate ft lne to [ O / H ] 0. [ He

More information

Difference Equations

Difference Equations Dfference Equatons c Jan Vrbk 1 Bascs Suppose a sequence of numbers, say a 0,a 1,a,a 3,... s defned by a certan general relatonshp between, say, three consecutve values of the sequence, e.g. a + +3a +1

More information

Feb 14: Spatial analysis of data fields

Feb 14: Spatial analysis of data fields Feb 4: Spatal analyss of data felds Mappng rregularly sampled data onto a regular grd Many analyss technques for geophyscal data requre the data be located at regular ntervals n space and/or tme. hs s

More information

THE ROYAL STATISTICAL SOCIETY 2006 EXAMINATIONS SOLUTIONS HIGHER CERTIFICATE

THE ROYAL STATISTICAL SOCIETY 2006 EXAMINATIONS SOLUTIONS HIGHER CERTIFICATE THE ROYAL STATISTICAL SOCIETY 6 EXAMINATIONS SOLUTIONS HIGHER CERTIFICATE PAPER I STATISTICAL THEORY The Socety provdes these solutons to assst canddates preparng for the eamnatons n future years and for

More information

C4B Machine Learning Answers II. = σ(z) (1 σ(z)) 1 1 e z. e z = σ(1 σ) (1 + e z )

C4B Machine Learning Answers II. = σ(z) (1 σ(z)) 1 1 e z. e z = σ(1 σ) (1 + e z ) C4B Machne Learnng Answers II.(a) Show that for the logstc sgmod functon dσ(z) dz = σ(z) ( σ(z)) A. Zsserman, Hlary Term 20 Start from the defnton of σ(z) Note that Then σ(z) = σ = dσ(z) dz = + e z e z

More information

2.3 Nilpotent endomorphisms

2.3 Nilpotent endomorphisms s a block dagonal matrx, wth A Mat dm U (C) In fact, we can assume that B = B 1 B k, wth B an ordered bass of U, and that A = [f U ] B, where f U : U U s the restrcton of f to U 40 23 Nlpotent endomorphsms

More information

Example: (13320, 22140) =? Solution #1: The divisors of are 1, 2, 3, 4, 5, 6, 9, 10, 12, 15, 18, 20, 27, 30, 36, 41,

Example: (13320, 22140) =? Solution #1: The divisors of are 1, 2, 3, 4, 5, 6, 9, 10, 12, 15, 18, 20, 27, 30, 36, 41, The greatest common dvsor of two ntegers a and b (not both zero) s the largest nteger whch s a common factor of both a and b. We denote ths number by gcd(a, b), or smply (a, b) when there s no confuson

More information

x = , so that calculated

x = , so that calculated Stat 4, secton Sngle Factor ANOVA notes by Tm Plachowsk n chapter 8 we conducted hypothess tests n whch we compared a sngle sample s mean or proporton to some hypotheszed value Chapter 9 expanded ths to

More information

The Gaussian classifier. Nuno Vasconcelos ECE Department, UCSD

The Gaussian classifier. Nuno Vasconcelos ECE Department, UCSD he Gaussan classfer Nuno Vasconcelos ECE Department, UCSD Bayesan decson theory recall that we have state of the world X observatons g decson functon L[g,y] loss of predctng y wth g Bayes decson rule s

More information

U.C. Berkeley CS294: Spectral Methods and Expanders Handout 8 Luca Trevisan February 17, 2016

U.C. Berkeley CS294: Spectral Methods and Expanders Handout 8 Luca Trevisan February 17, 2016 U.C. Berkeley CS94: Spectral Methods and Expanders Handout 8 Luca Trevsan February 7, 06 Lecture 8: Spectral Algorthms Wrap-up In whch we talk about even more generalzatons of Cheeger s nequaltes, and

More information

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur Analyss of Varance and Desgn of Exerments-I MODULE III LECTURE - 2 EXPERIMENTAL DESIGN MODELS Dr. Shalabh Deartment of Mathematcs and Statstcs Indan Insttute of Technology Kanur 2 We consder the models

More information

Instance-Based Learning (a.k.a. memory-based learning) Part I: Nearest Neighbor Classification

Instance-Based Learning (a.k.a. memory-based learning) Part I: Nearest Neighbor Classification Instance-Based earnng (a.k.a. memory-based learnng) Part I: Nearest Neghbor Classfcaton Note to other teachers and users of these sldes. Andrew would be delghted f you found ths source materal useful n

More information

C/CS/Phy191 Problem Set 3 Solutions Out: Oct 1, 2008., where ( 00. ), so the overall state of the system is ) ( ( ( ( 00 ± 11 ), Φ ± = 1

C/CS/Phy191 Problem Set 3 Solutions Out: Oct 1, 2008., where ( 00. ), so the overall state of the system is ) ( ( ( ( 00 ± 11 ), Φ ± = 1 C/CS/Phy9 Problem Set 3 Solutons Out: Oct, 8 Suppose you have two qubts n some arbtrary entangled state ψ You apply the teleportaton protocol to each of the qubts separately What s the resultng state obtaned

More information

CS 3710: Visual Recognition Classification and Detection. Adriana Kovashka Department of Computer Science January 13, 2015

CS 3710: Visual Recognition Classification and Detection. Adriana Kovashka Department of Computer Science January 13, 2015 CS 3710: Vsual Recognton Classfcaton and Detecton Adrana Kovashka Department of Computer Scence January 13, 2015 Plan for Today Vsual recognton bascs part 2: Classfcaton and detecton Adrana s research

More information

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE Analytcal soluton s usually not possble when exctaton vares arbtrarly wth tme or f the system s nonlnear. Such problems can be solved by numercal tmesteppng

More information

Chapter 2 - The Simple Linear Regression Model S =0. e i is a random error. S β2 β. This is a minimization problem. Solution is a calculus exercise.

Chapter 2 - The Simple Linear Regression Model S =0. e i is a random error. S β2 β. This is a minimization problem. Solution is a calculus exercise. Chapter - The Smple Lnear Regresson Model The lnear regresson equaton s: where y + = β + β e for =,..., y and are observable varables e s a random error How can an estmaton rule be constructed for the

More information

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers Psychology 282 Lecture #24 Outlne Regresson Dagnostcs: Outlers In an earler lecture we studed the statstcal assumptons underlyng the regresson model, ncludng the followng ponts: Formal statement of assumptons.

More information

Affine transformations and convexity

Affine transformations and convexity Affne transformatons and convexty The purpose of ths document s to prove some basc propertes of affne transformatons nvolvng convex sets. Here are a few onlne references for background nformaton: http://math.ucr.edu/

More information

OPTIMISATION. Introduction Single Variable Unconstrained Optimisation Multivariable Unconstrained Optimisation Linear Programming

OPTIMISATION. Introduction Single Variable Unconstrained Optimisation Multivariable Unconstrained Optimisation Linear Programming OPTIMIATION Introducton ngle Varable Unconstraned Optmsaton Multvarable Unconstraned Optmsaton Lnear Programmng Chapter Optmsaton /. Introducton In an engneerng analss, sometmes etremtes, ether mnmum or

More information

A Bayes Algorithm for the Multitask Pattern Recognition Problem Direct Approach

A Bayes Algorithm for the Multitask Pattern Recognition Problem Direct Approach A Bayes Algorthm for the Multtask Pattern Recognton Problem Drect Approach Edward Puchala Wroclaw Unversty of Technology, Char of Systems and Computer etworks, Wybrzeze Wyspanskego 7, 50-370 Wroclaw, Poland

More information

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur Analyss of Varance and Desgn of Experment-I MODULE VII LECTURE - 3 ANALYSIS OF COVARIANCE Dr Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur Any scentfc experment s performed

More information

The Order Relation and Trace Inequalities for. Hermitian Operators

The Order Relation and Trace Inequalities for. Hermitian Operators Internatonal Mathematcal Forum, Vol 3, 08, no, 507-57 HIKARI Ltd, wwwm-hkarcom https://doorg/0988/mf088055 The Order Relaton and Trace Inequaltes for Hermtan Operators Y Huang School of Informaton Scence

More information

8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS

8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS SECTION 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS 493 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS All the vector spaces you have studed thus far n the text are real vector spaces because the scalars

More information

Lossy Compression. Compromise accuracy of reconstruction for increased compression.

Lossy Compression. Compromise accuracy of reconstruction for increased compression. Lossy Compresson Compromse accuracy of reconstructon for ncreased compresson. The reconstructon s usually vsbly ndstngushable from the orgnal mage. Typcally, one can get up to 0:1 compresson wth almost

More information

Resource Allocation and Decision Analysis (ECON 8010) Spring 2014 Foundations of Regression Analysis

Resource Allocation and Decision Analysis (ECON 8010) Spring 2014 Foundations of Regression Analysis Resource Allocaton and Decson Analss (ECON 800) Sprng 04 Foundatons of Regresson Analss Readng: Regresson Analss (ECON 800 Coursepak, Page 3) Defntons and Concepts: Regresson Analss statstcal technques

More information

Chapter 12 Analysis of Covariance

Chapter 12 Analysis of Covariance Chapter Analyss of Covarance Any scentfc experment s performed to know somethng that s unknown about a group of treatments and to test certan hypothess about the correspondng treatment effect When varablty

More information

A Novel Biometric Feature Extraction Algorithm using Two Dimensional Fisherface in 2DPCA subspace for Face Recognition

A Novel Biometric Feature Extraction Algorithm using Two Dimensional Fisherface in 2DPCA subspace for Face Recognition A Novel ometrc Feature Extracton Algorthm usng wo Dmensonal Fsherface n 2DPA subspace for Face Recognton R. M. MUELO, W.L. WOO, and S.S. DLAY School of Electrcal, Electronc and omputer Engneerng Unversty

More information

5 The Rational Canonical Form

5 The Rational Canonical Form 5 The Ratonal Canoncal Form Here p s a monc rreducble factor of the mnmum polynomal m T and s not necessarly of degree one Let F p denote the feld constructed earler n the course, consstng of all matrces

More information

p 1 c 2 + p 2 c 2 + p 3 c p m c 2

p 1 c 2 + p 2 c 2 + p 3 c p m c 2 Where to put a faclty? Gven locatons p 1,..., p m n R n of m houses, want to choose a locaton c n R n for the fre staton. Want c to be as close as possble to all the house. We know how to measure dstance

More information

Section 8.3 Polar Form of Complex Numbers

Section 8.3 Polar Form of Complex Numbers 80 Chapter 8 Secton 8 Polar Form of Complex Numbers From prevous classes, you may have encountered magnary numbers the square roots of negatve numbers and, more generally, complex numbers whch are the

More information

The Multiple Classical Linear Regression Model (CLRM): Specification and Assumptions. 1. Introduction

The Multiple Classical Linear Regression Model (CLRM): Specification and Assumptions. 1. Introduction ECONOMICS 5* -- NOTE (Summary) ECON 5* -- NOTE The Multple Classcal Lnear Regresson Model (CLRM): Specfcaton and Assumptons. Introducton CLRM stands for the Classcal Lnear Regresson Model. The CLRM s also

More information

Lecture Nov

Lecture Nov Lecture 18 Nov 07 2008 Revew Clusterng Groupng smlar obects nto clusters Herarchcal clusterng Agglomeratve approach (HAC: teratvely merge smlar clusters Dfferent lnkage algorthms for computng dstances

More information

Mathematical Preparations

Mathematical Preparations 1 Introducton Mathematcal Preparatons The theory of relatvty was developed to explan experments whch studed the propagaton of electromagnetc radaton n movng coordnate systems. Wthn expermental error the

More information

Homework Assignment 3 Due in class, Thursday October 15

Homework Assignment 3 Due in class, Thursday October 15 Homework Assgnment 3 Due n class, Thursday October 15 SDS 383C Statstcal Modelng I 1 Rdge regresson and Lasso 1. Get the Prostrate cancer data from http://statweb.stanford.edu/~tbs/elemstatlearn/ datasets/prostate.data.

More information

Chapter 9: Statistical Inference and the Relationship between Two Variables

Chapter 9: Statistical Inference and the Relationship between Two Variables Chapter 9: Statstcal Inference and the Relatonshp between Two Varables Key Words The Regresson Model The Sample Regresson Equaton The Pearson Correlaton Coeffcent Learnng Outcomes After studyng ths chapter,

More information

Chapter 14 Simple Linear Regression

Chapter 14 Simple Linear Regression Chapter 4 Smple Lnear Regresson Chapter 4 - Smple Lnear Regresson Manageral decsons often are based on the relatonshp between two or more varables. Regresson analss can be used to develop an equaton showng

More information

Lecture 6: Introduction to Linear Regression

Lecture 6: Introduction to Linear Regression Lecture 6: Introducton to Lnear Regresson An Manchakul amancha@jhsph.edu 24 Aprl 27 Lnear regresson: man dea Lnear regresson can be used to study an outcome as a lnear functon of a predctor Example: 6

More information

Problem Set 9 Solutions

Problem Set 9 Solutions Desgn and Analyss of Algorthms May 4, 2015 Massachusetts Insttute of Technology 6.046J/18.410J Profs. Erk Demane, Srn Devadas, and Nancy Lynch Problem Set 9 Solutons Problem Set 9 Solutons Ths problem

More information

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal Inner Product Defnton 1 () A Eucldean space s a fnte-dmensonal vector space over the reals R, wth an nner product,. Defnton 2 (Inner Product) An nner product, on a real vector space X s a symmetrc, blnear,

More information

Statistics for Managers Using Microsoft Excel/SPSS Chapter 13 The Simple Linear Regression Model and Correlation

Statistics for Managers Using Microsoft Excel/SPSS Chapter 13 The Simple Linear Regression Model and Correlation Statstcs for Managers Usng Mcrosoft Excel/SPSS Chapter 13 The Smple Lnear Regresson Model and Correlaton 1999 Prentce-Hall, Inc. Chap. 13-1 Chapter Topcs Types of Regresson Models Determnng the Smple Lnear

More information

A linear imaging system with white additive Gaussian noise on the observed data is modeled as follows:

A linear imaging system with white additive Gaussian noise on the observed data is modeled as follows: Supplementary Note Mathematcal bacground A lnear magng system wth whte addtve Gaussan nose on the observed data s modeled as follows: X = R ϕ V + G, () where X R are the expermental, two-dmensonal proecton

More information

Supporting Information

Supporting Information Supportng Informaton The neural network f n Eq. 1 s gven by: f x l = ReLU W atom x l + b atom, 2 where ReLU s the element-wse rectfed lnear unt, 21.e., ReLUx = max0, x, W atom R d d s the weght matrx to

More information

Correlation and Regression. Correlation 9.1. Correlation. Chapter 9

Correlation and Regression. Correlation 9.1. Correlation. Chapter 9 Chapter 9 Correlaton and Regresson 9. Correlaton Correlaton A correlaton s a relatonshp between two varables. The data can be represented b the ordered pars (, ) where s the ndependent (or eplanator) varable,

More information