Projection onto A Nonnegative Max-Heap

Size: px
Start display at page:

Download "Projection onto A Nonnegative Max-Heap"

Transcription

1 Projection onto A Nonnegatie Max-Heap Jun Liu Arizona State Uniersity Tempe, AZ 85287, USA j.iu@asu.edu Liang Sun Arizona State Uniersity Tempe, AZ 85287, USA sun.iang@asu.edu Jieping Ye Arizona State Uniersity Tempe, AZ 85287, USA jieping.ye@asu.edu Abstract We consider the probem of computing the Eucidean projection of a ector of ength p onto a non-negatie max-heap an ordered tree where the aues of the nodes are a nonnegatie and the aue of any parent node is no ess than the aue(s) of its chid node(s). This Eucidean projection pays a buiding bock roe in the optimization probem with a non-negatie maxheap constraint. Such a constraint is desirabe when the features foow an ordered tree structure, that is, a gien feature is seected for the gien regression/cassification task ony if its parent node is seected. In this paper, we show that such Eucidean projection probem admits an anaytica soution and we deeop a top-down agorithm where the key operation is to find the so-caed maxima root-tree of the subtree rooted at each node. A naie approach for finding the maxima root-tree is to enumerate a the possibe root-trees, which, howeer, does not scae we. We reea seera important properties of the maxima root-tree, based on which we design a bottom-up agorithm with merge for efficienty finding the maxima roottree. The proposed agorithm has a (worst-case) inear time compexity for a sequentia ist, and O(p 2 ) for a genera tree. We report simuation resuts showing the effectieness of the max-heap for regression with an ordered tree structure. Empirica resuts show that the proposed agorithm has an expected inear time compexity for many specia cases incuding a sequentia ist, a fu binary tree, and a tree with depth. Introduction In many regression/cassification probems, the features exhibit certain hierarchica or structura reationships, the usage of which can yied an interpretabe mode with improed regression/cassification performance [25]. Recenty, there hae been increasing interests on structured sparisty with arious approaches for incorporating structures; see [7, 8, 9, 7, 24, 25] and references therein. In this paper, we consider an ordered tree structure: a gien feature is seected for the gien regression/cassification task ony if its parent node is seected. To incorporate such ordered tree structure, we assume that the mode parameter x R p foows the non-negatie max-heap structure : P = {x,x i x j (x i,x j ) E t }, () where T t = (V t,e t ) is a target tree with V t = {x,x 2,...,x p } containing a the nodes and E t a the edges. The constraint set P impies that if x i is the parent node of a chid node x j then the aue of x i is no ess than the aue of x j. In other words, if a parent node x i is, then any of its chid nodes x j is aso. Figure iustrates three specia tree structures: ) a fu binary tree, 2) a sequentia ist, and 3) a tree with depth. To dea with the negatie mode parameters, one can make use of the technique empoyed in [24], which soes the scaed ersion of the east square estimate.

2 x x x2 x3 x x2 x3 x4 x5 x6 x7 x2 x3 x4 x5 x6 x7 x4 x5 x6 x7 (a) (b) (c) Figure : Iustration of a non-negatie max-heap depicted in(). Pots(a), (b), and(c) correspond to a fu binary tree, a sequentia ist, and a tree with depth, respectiey. The set P defined in () induces the so-caed heredity principe [3, 6, 8, 24], which has been proen effectie for high-dimensiona ariabe seection. In a recent study [2], Li et a. conducted a meta-anaysis of 3 data sets from pubished factoria experiments and concuded that an oerwheming majority of these rea studies conform with the heredity principes. The ordered tree structure is a specia case of the non-negatie garrote discussed in [24] when the hierarchica reationship is depicted by a tree. Therefore, the asymptotic properties estabished in [24] are appicabe to the ordered tree structrue. Seera reated approaches that can incorporate the ordered tree structure incude the Wedge approach [7] and the hierarchica group Lasso [25]. The Wedge approach incorporates such ordering information p i= (x2 i t i +t i ), with by designing a penaty for the mode parameter x as Ω(x P) = inf t P 2 tree being a sequentia ist. By imposing the mixed - 2 norm on each group formed by the nodes in the subtree of a parent node, the hierarchica group Lasso is abe to incorporate such ordering information. The hierarchica group Lasso has been appied for muti-task earning in [] with a tree structure, and the efficient computation was discussed in [, 5]. Compared to Wedge and hierarchica group Lasso, the max-heap in () incorporates such ordering information in a direct way, and our simuation resuts show that the max-heap can achiee ower reconstruction error than both approaches. In estimating the mode parameter satisfying the ordered tree structure, one needs to soe the foowing constrained optimization probem: minf(x) (2) x P for some conex function f( ). The probem(2) can be soed ia many approaches incuding subgradient descent, cutting pane method, gradient descent, acceerated gradient descent, etc [9, 2]. In appying these approaches, a key buiding bock is the so-caed Eucidean projection of a ector onto the conex set P: π P () = argmin x P 2 x 2 2, (3) which ensures that the soution beongs to the constraint set P. For some specia set P (e.g., hyperpane, hafspace, and rectange), the Eucidean projection admits a simpe anaytica soution, see [2]. In the iterature, researchers hae deeoped efficient Eucidean projection agorithms for the -ba [5, 4], the / 2 -ba [], and the poyhedra [4, 22]. When P is induced by a sequentia ist, a inear time agorithm was recenty proposed in [26]. Without the non-negatie constraints, probem(3) is the so-caed isotonic regression probem[6, 2]. Our major technica contribution in this paper is the efficient computation of (3) for the set P defined in (). In Section 2, we show that the Eucidean projection admits an anaytica soution, and we deeop a top-down agorithm where the key operation is to find the so-caed maxima root-tree of the subtree rooted at each node. In Section 3, we design a bottom-up agorithm with merge for efficienty finding the maxima root-tree by using its properties. We proide empirica resuts for the proposed agorithm in Section 4, and concude this paper in Section 5. 2 Atda: A Top-Down Agorithm In this section, we deeop an agorithm in a top-down manner caed Atda for soing (3). With the target tree T t = (V t,e t ), we construct the input tree T = (V,E) with the input ector, where V = {, 2,..., p } and E = {( i, j ) (x i,x j ) E t }. For the conenience of presenting our proposed agorithm, we begin with seera definitions. We aso proide some exampes for eaborating the definitions in the suppementary fie A.. 2

3 Definition. For a non-empty tree T = (V,E), we define its root-tree as any non-empty tree T = (Ṽ,Ẽ) that satisfies: ) Ṽ V, 2) Ẽ E, and 3) T shares the same root as T. Definition 2. For a non-empty tree T = (V,E), we define R(T) as the root-tree set containing a its root-trees. Definition 3. For a non-empty tree T = (V,E), we define ( m(t) = max ) i V i,, (4) V which equas the mean of a the nodes in T if such mean is non-negatie, and otherwise. Definition 4. For a non-empty tree T = (V,E), we define its maxima root-tree as: where M max (T) = arg max T=(Ṽ,Ẽ): T R(T),m( T)=m max(t) Ṽ, (5) m max (T) = max m( T) (6) T R(T) is the maxima aue of a the root-trees of the tree T. Note that if two root-trees share the same maxima aue, (5) seects the one with the argest tree size. When T = (Ṽ,Ẽ) is a part of a arger tree T = (V,E), i.e., Ṽ V and Ẽ E, we can treat T as a super-node of the tree T with aue m( T). Thus, we hae the foowing definition of a super-tree (note that a super-tree proides a disjoint partition of the gien tree): Definition 5. For a non-empty tree T = (V,E), we define its super-tree as S = (V S,E S ), which satisfies: ) each node in V S = {T,T 2,...,T n } is a non-empty tree with T i = (V i,e i ), 2) V i V and E i E, 3) V i Vj =,i j and V = n i= V i, and 4) (T i,t j ) E S if and ony if there exists a node in T j whose parent node is in T i. 2. Proposed Agorithm We present the pseudo code for soing (3) in Agorithm. The key idea of the proposed agorithm is that, in the i-th ca, we find T i = M max (T), the maxima root-tree of T, set x corresponding to the nodes of T i to m i = m max (T) = m(t i ), remoe T i from the tree T, and appy Atda to the resuting trees one by one recursiey. Agorithm A Top-Down Agorithm: Atda Input: the tree structure T = (V,E), i Output: x R p : Set i = i+ 2: Find the maxima root-tree of T, denoted by T i = (V i,e i ), and set m i = m(t i ) 3: if m i > then 4: Set x j = m i, j V i 5: Remoe the root-tree T i from T, denote the resuting trees as T, T 2,..., T ri, and appy Atda( T j,i), j =,2,...,r i 6: ese 7: Set x j = m i, j V i 8: end if 2.2 Iustration & Justification For a better iustration and justification of the proposed agorithm, we proide the anaysis of Atda for a specia case the sequentia ist in the suppementary fie A.2. Let us anayze Agorithm for the genera tree. Figure 2 iustrates soing (3) ia Agorithm for a tree with depth 3. Pot (a) shows a target tree T t, and pot (b) denotes the input tree T. The dashed frame of pot (b) shows M max (T), the maxima root-tree of T, and 3

4 x x2 x3 x x5 x6 x7 x8 x9 x x x2 x3 x4 x5 (a) (b) (c) (f) (e) Figure 2: Iustration of Agorithm for soing (3) for a tree with depth 3. Pot (a) shows the target tree T t, and pots (b-e) iustrate Atda. Specificay, pot (b) denotes the input tree T, with the dashed frame dispaying its maxima root-tree; pot (c) depicts the resuting trees after remoing the maxima root-tree in pot (b); pot (d) shows the resuting super-tree (we treat each tree encosed by the dashed frame as a super-node) of the agorithm; pot (e) gies the soution x R 5 ; and the edges of pot (f) show the dua ariabes, from which we can aso obtain the optima soution x (refer to the proof of Theorem ). (d) we hae M max (T) = 3. Thus, we set the corresponding entries of x to 3. Pot (c) depicts the resuting trees after remoing the maxima root-tree in pot (b), and pot (d) shows the generated maxima root-trees (encosed by dashed frame) by the agorithm. When treating each generated maxima root-tree as a super-node with the aue defined in Definition 3, pot (d) is a super-tree of the input tree T. In addition, the super-tree is a max-heap, i.e., the aue of the parent node is no ess than the aues of its chid nodes. Pot (e) gies the soution x R 5. The edges of pot (f) correspond to the aues of the dua ariabes, from which we can aso obtain the optima soution x R 5. Finay, we can obsere that the non-zero entries of x constitute a cut of the origina tree. We erify the correctness of Agorithm for the genera tree in the foowing theorem. We make use of the KKT conditions and ariationa inequaity [2] in the proof. Theorem. x = Atda(T, ) proides the unique optima soution to (3). Proof: As the objectie function of (3) is stricty conex and the constraints are affine, it admits a unique soution. After running Agorithm, we obtain the sequences {T i } k i= and {m i } k i=, where k satisfies k p. It is easy to erify that the trees T i,i =,2,...,k constitute a disjoint partition of the input tree T. With the sequences {T i } k i= and {m i} k i=, we can construct a super-tree of the input tree T as foows: ) we treat T i as a super-node with aue m i, and 2) we put an edge between T i and T j if there is an edge between the nodes of T i and T j in the input tree T. With Agorithm, we can erify that the resuting super-tree has the property that the aue of the parent node is no ess than its chid nodes. Therefore, x = Atda(T,) satisfies x P. Let x and denote a subset of x and corresponding to the indices appearing in the subtree T, respectiey. Denote P = {x : x,x i x j,( i, j ) E }, I = { : m > },I 2 = { : m = }. Our proof is based on the foowing inequaity: min x P 2 x 2 2 min x I P 2 x min x I P 2 x 2 2, (7) 2 which hods as the eft hand side has the additiona inequaity constraints compared to the right hand side. Our methodoogy is to show that x = Atda(T,) proides the optima soution to the right hand side of (7), i.e., x = arg min x P 2 x 2 2, I, (8) x = arg min x P 2 x 2 2, I 2, (9) 4

5 which, together with the fact 2 x 2 2 min x P 2 x 2 2, x P ead to our main argument. Next, we proe (8) by the KKT conditions, and proe (9) by the ariationa inequaity [2]. Firsty, I, we introduce the dua ariabe y ij for the edge ( i, j ) E, and y ii if i L, where L contains a the eaf nodes of the tree T. Denote the root of T by r. For a i V, i r, we denote its parent node by ji, and for the root r, we denote j r = r. We et C i = {j j is a chid node of i in the tree T }. R i = {j j is in the subtree of T rooted at i }. To proe (8), we erify that the prima ariabe x = Atda(T,) and the dua ariabe ỹ satisfy the foowing KKT conditions: ( i, j ) E, x i x j () ( i, j ) E,( x i x j )ỹ ij = () i L,ỹ ii x i = (2) i V, x i i ỹ ij +ỹ jii = (3) j C i ( i, j ) E,ỹ ij (4) i L,ỹ ii, (5) where ỹ jr r = (Note that ỹ jr r is a dua ariabe, and it is introduced for the simpicity of presenting (2)), and the dua ariabe ỹ is set as: ỹ ii =, i L, (6) ỹ jii = i m + ỹ ij, i V. (7) According to Agorithm, x i = m >, i V, I. Thus, we hae ()-(2) and (5). It foows from (7) that (3) hods. According to (6) and (7), we hae ỹ jii = j Ri m, i V, (8) j R i j C i where Ri denotes the number of eements in R i, the subtree of T rooted at i. From the nature of the maxima root-tree T, I, we hae j R i j Ri m. Otherwise, if j R i j < Ri m, we can construct from T a new root-tree T by remoing the subtree of T rooted at i, so that T achiees a arger aue than T. This contradicts with the argument that T, I is the maxima root-tree of the working tree T. Therefore, it foows from (8) that (4) hods. Secondy, we proe (9) by erifying the foowing optimaity condition: x x, x, x P, I 2, (9) whichistheso-caedariationainequaityconditionfor x beingtheoptimasoutionto(9). According to Agorithm, if I 2, we hae x i =, i V. Thus, (9) is equiaent to x,, x P, I 2. (2) For a gien x P, if x i =, i V, (2) naturay hods. Next, we consider x. Denote by x the minima nonzero eement in x, and T = (V,E ) a tree constructed by remoing the nodes corresponding to the indices in the set {i : x i =, i V } from T. It is cear that T shares the same root as T. It foows from Agorithm that i: i V i. Thus, we hae x, = x i + (x i x ) i (x i x ) i. i: i V i: i V 5 i: i V

6 If x i = x, i V, we arrie at (2). Otherwise, we set r = 2; denote by x r the minima nonzero eement in the set {x i r j= x j : i V r }, and T r = (V r,er ) a subtree of T r by remoing those nodes with the indices in the set {i : x i r j= x j =, i V r }. It is cear that T r shares the same root as T r and T as we, so that it foows from Agorithm that i: i V r i. Therefore, we hae r (x i x j) i = x r i + r (x i x j) i r (x i x j) i. (2) i: i V r j= i: i V r i: i V r j= i: i V r Repeating the aboe process unti V r is empty, we can erify that (2) hods. For a better understanding of the proof, we make use of the edges of Figure 2 (f) to show the dua ariabes, where the edge connecting i and j corresponds to the dua ariabe ỹ ij, and the edge starting from the eaf node i corresponds to the dua ariabe ỹ ii. With the dua ariabes, we can compute x ia (3). We note that, for the maxima root-tree with a positie aue, the associated dua ariabes are unique, but for the maxima root-tree with zero aue, the associated dua ariabes may not be unique. For exampe, in Figure 2 (f), we set ỹ ii = for i = 2, ỹ ii = for i = 3, ỹ ij = 2 for i = 6,j = 2, and ỹ ij = 2 for i = 6,j = 3. It is easy to check that the dua ariabes can aso be set as foows: ỹ ii = for i = 2, ỹ ii = for i = 3, ỹ ij = for i = 6,j = 2, and ỹ ij = 3 for i = 6,j = 3. 3 Finding the Maxima Root-Tree A key operation of Agorithm is to find the maxima root-tree used in Step 2. A naie approach for finding the maxima root-tree of a tree T is to enumerate a possibe roottrees in the root-tree set R(T), and identify the maxima root-tree ia (5). We ca such an approach Anae, which stands for a naie agorithm with enumeration. Athough Anae is simpe to describe, it has a ery high time compexity (see the anaysis gien in suppementary fie A.3). To this end, we deeop Abuam (A Bottom-Up Agorithm with Merge). The underying idea is to make use of the specia structure of the maxima root-tree defined in (5) for aoiding the enumeration of a possibe root-trees. We begin the discussion with some key properties of the maxima root-tree, and the proof is gien in the suppementary fie A.4. Lemma. For a non-empty tree T = (V,E), denote its maxima root-tree as T max = (V max,e max ). Let T = ( Ṽ,Ẽ) be a root-tree of T max. Assume that there are n nodes i,..., in, which satisfy: ) ij / Ṽ, 2) i j V, and 3) the parent node of ij is in Ṽ. If n, we denote the subtree of T rooted at ij as T j = (V j,e j ),j =,2,...,n, T j max = (V j max,e j max) as the maxima root-trees of T j, and m = max j=,2,...,n m(t j max). Then, the foowings hod: () If n =, then T max = T = T; (2) If n, m( T) =, and m =, then T max = T; (3) If n, m( T) >, and m( T) > m, then T max = T; (4) If n, m( T) >, and m( T) m, then V j max V max, E j max E max and ( i, ij ) E max, j : m(t j max) = m; and (5) If n, m( T) =, and m >, then V j max V max, E j max E max and ( i, ij ) E max, j : m(t j max) = m. For the conenience of presenting our proposed agorithm, we define the operation merge as foows: Definition 6. Let T = (V,E) be a non-empty tree, and T = (V,E ) and T 2 = (V 2,E 2 ) be two trees that satisfy: ) they are composed of a subset of the nodes and edges of T, i.e., V V, V 2 V, E E, and E 2 E; 2) they do not oerap, i.e., V V 2 =, and E E 2 = ; and 3) in the tree T, i2, the root node of T 2 is a chid of i, a eaf node of T. We define the operation merge as T = merge(t,t 2,T), where T = (Ṽ,Ẽ) with V = V V2 and E = E E2 {(i, i2 )}. Next, we make use of Lemma to efficienty compute the maxima root-tree, and present the pseudo code for Abuam in Agorithm 2. We proide the iustration of the proposed agorithm and the anaysis of its computationa cost in the suppementary fie A.5 and A.6, respectiey. j= 6

7 Agorithm 2 A Bottom-Up Agorithm with Merge: Abuam Input: the input tree T = (V,E) Output: the maxima root-tree T max = (V max,e max ) : Set T = (V,E ), where V = {x i } and E = 2: if i does not hae a chid node in T then 3: Set T max = T, return 4: end if 5: whie do 6: Set m =, denote by i,..., in the n nodes that satisfy: ) ij / V, 2) ij V, and 3) the parent node of ij is in V, and denote by T j = (V j,e j ),j =,2,...,n the subtree of T rooted at ij. 7: if n = then 8: Set T max = T = T, return 9: end if : for j = to n do : Set T j max = Abuam(T j ), and m = max(m(t j max), m) 2: end for 3: if m(t ) = m = then 4: Set T max = T, return 5: ese if m( T) > and m( T) > m then 6: Set T max = T, return 7: ese 8: Set T =merge(t, T j max, T), j : m(t j max) = m 9: end if 2: end whie Making use of the fact that T is aways a aid root-tree of T max, the maxima root-tree of T, we can easiy proe the foowing resut using Lemma. Theorem 2. T max returned by Agorithm 2 is the maxima root-tree of the input tree T. 4 Numerica Simuations Effectieness of the Max-Heap Structure We test the effectieness of the max-heap structure for inear regression b = Ax, foowing the same experimenta setting as in [7]. Specificay, the eements of A R n p are generated i.i.d. from the Gaussian distribution with zero mean and standard deriation and the coumns of A are then normaized to hae unit ength. The regression ector x has p = 27 nonincreasing eements, where the first eements are set as x i = i,i =,2,..., and the rest are zeros. We compared with the foowing three approaches: Lasso [23], Group Lasso [25], and Wedge [7]. Lasso makes no use of such ordering, whie Wedge incorporates the structure by using an auxiiary ordered ariabe. For Group Lasso and Max-Heap, we try binary-tree grouping and ist-tree grouping, where the associated trees are a fu binary tree and a sequentia ist, respectiey. The regression ector is put on the tree so that, the coser the node to the root, the arger the eement is paced. In Group Lasso, the nodes appearing in the same subtree form a group. For the compared approaches, we use the impementations proided in [7] 2 ; and for Max-Heap, we soe (2) with f(x) = 2 Ax b 2 2+ρ x for some sma ρ = r A T b (we set r = 4, and 8 for the binary-tree grouping and ist-tree grouping, respectiey) and appy the acceerated gradient descent[9] approach with our proposed Eucidean projection. We compute the aerage mode error x x 2 oer 5 independent runs, and report the resutswithaaryingnumberofsampesizeninfigure3(a)&(b). Asexpected, GL-binary, MH-binary, Wedge, GL-ist and MH-ist outperform Lasso which does not incorporate such ordering information. MH-binary performs better than GL-binary, and MH-ist performs better than Wedge and GL-ist, due to the direct usage of such ordering information. In addition, the ist-tree grouping performs better than the binary-tree grouping, as it makes better usage of such ordering information

8 Lasso GL binary MH binary Gaussian Distribution for sequentia ist fu binary tree tree of depth Gaussian Distribution, Fu Binary Tree Mode error Computationa Time Computationa Time 2 3 d= d=2 d=4 d=8 d=8 d=2 Mode error Sampe size p Random Initiaization of (a) (c) (e) Wedge GL ist MH ist Computationa Time sequentia ist fu binary tree tree of depth Uniform Distribution for Computationa Time 2 3 Uniform Distribution, Fu Binary Tree d= d=2 d=4 d=8 d=8 d= Sampe size p Random Initiaization of (b) (d) (f) Figure 3: Simuation resuts. In pots (a) and (b), we show the aerage mode error x x 2 oer 5 independet runs by different approaches with the fu binary-tree ordering and the ist-tree ordering. In pots (c) and (d), we report the computationa time (in seconds) of the proposed Atda (aeraged oer runs) with different randomy initiaized input. In pots (e) and (f), we show the computationa time of Atda oer runs. Efficiency of the Proposed Projection We test the efficiency of the proposed Atda approach for soing the Eucidean projection onto the non-negatie max-heap, equipped with our proposed Abuam approach for finding the maxima root-trees. In the experiments, we make use of the three tree structures as depicted in Figure, and try two different distributions: ) Gaussian distribution with zero mean and standard deriation and 2) uniform distribution in [, ] for randomy and independenty generating the entries of the input R p. In Figure 3 (c) & (d), we report the aerage computationa time (in seconds) oer runs under different aues of p = 2 d+, where d =,2,...,2. We can obsere that, the proposed agorithm scaes ineary with the size of p. In Figure 3 (e) & (f), we report the computationa time of Atda oer runs when the ordered tree structure is a fu binary tree. The resuts show that the computationa time of the proposed agorithm is reatiey stabe for different runs, especiay for arger d or p. Note that, the source codes for our proposed agorithm hae been incuded in the SLEP package [3]. 5 Concusion In this paper, we hae deeoped an efficient agorithm for the computation of the Eucidean projection onto a non-negatie max-heap. The proposed agorithm has a (worst-case) inear time compexity for a sequentia ist, and O(p 2 ) for a genera tree. Empirica resuts show that: ) the proposed approach deas with the ordering information better than existing approaches, and 2) the proposed agorithm has an expected inear time compexity for the sequentia ist, the fu binary tree, and the tree of depth. It wi be interesting to expore whether the proposed Abuam has a worst case inear (or inearithmic) time compexity for the binary tree. We pan to appy the proposed agorithms to rea-word appications with an ordered tree structure. We aso pan to extend our proposed approaches to the genera hierarchica structure. Acknowedgments This work was supported by NSF IIS-8255, IIS , MCB-267, CCF-2577, NGA HM , and NSFC 69535,

9 References [] E. Berg, M. Schmidt, M. P. Friedander, and K. Murphy. Group sparsity ia inear-time projection. Tech. Rep. TR-28-9, Department of Computer Science, Uniersity of British Coumbia, Vancouer, Juy 28. [2] S. Boyd and L. Vandenberghe. Conex Optimization. Cambridge Uniersity Press, 24. [3] N. Choi, W. Li, and J. Zhu. Variabe seection with the strong heredity constraint and its orace property. Journa of the American Statistica Association, 5: , 2. [4] Z. Dostá. Box constrained quadratic programming with proportioning and projections. SIAM Journa on Optimization, 7(3):87 887, 997. [5] J. Duchi, S. Shae-Shwartz, Y. Singer, and C. Tushar. Efficient projection onto the -ba for earning in high dimensions. In Internationa Conference on Machine Learning, 28. [6] M. Hamada and C. Wu. Anaysis of designed experiments with compex aiasing. Journa of Quaity Technoogy, 24:3 37, 992. [7] J. Huang, T. Zhang, and D. Metaxas. Learning with structured sparsity. In Internationa Conference on Machine Learning. 29. [8] L. Jacob, G. Obozinski, and J. Vert. Group asso with oerap and graph asso. In Internationa Conference on Machine Learning, 29. [9] R. Jenatton, J.-Y. Audibert, and F. Bach. Structured ariabe seection with sparsity-inducing norms. Technica report, arxi: , 29. [] R. Jenatton, J. Maira, G. Obozinski, and F. Bach. Proxima methods for sparse hierarchica dictionary earning. In Internationa Conference on Machine Learning, 2. [] S. Kim and E. P. Xing. Tree-guided group asso for muti-task regression with structured sparsity. In Internationa Conference on Machine Learning, 2. [2] X. Li, N. Sundarsanam, and D. Frey. Reguarities in data from factoria experiments. Compexity, :32 45, 26. [3] J. Liu, S. Ji, and J. Ye. SLEP: Sparse Learning with Efficient Projections. Arizona State Uniersity, 29. [4] J. Liu and J. Ye. Efficient Eucidean projections in inear time. In Internationa Conference on Machine Learning, 29. [5] J. Liu and J. Ye. Moreau-yosida reguarization for grouped tree structure earning. In Adances in Neura Information Processing Systems, 2. [6] R. Luss, S. Rosset, and M. Shahar. Decomposing isotonic regression for efficienty soing arge probems. In Adances in Neura Information Processing Systems, 2. [7] C. Micchei, J. Moraes, and M. Ponti. A famiy of penaty functions for structured sparsity. In Adances in Neura Information Processing Systems 23, pages [8] J. Neder. The seection of terms in response-surface modes how strong is the weak-heredity principe? Annas of Appied Statistics, 52:35 38, 998. [9] A. Nemiroski. Efficient methods in conex programming. Lecture Notes, 994. [2] Y. Nestero. Introductory Lectures on Conex Optimization: A Basic Course. Kuwer Academic Pubishers, 24. [2] P. M. Pardaos and G. Xue. Agorithms for a cass of isotonic regression probems. Agorithmica, 23:2 222, 999. [22] S. Shae-Shwartz and Y. Singer. Efficient earning of abe ranking by soft projections onto poyhedra. Journa of Machine Learning Research, 7: , 26. [23] R. Tibshirani. Regression shrinkage and seection ia the asso. Journa of the Roya Statistica Society Series B, 58(): , 996. [24] M. Yuan, V. R. Joseph, and H. Zou. Structured ariabe seection and estimation. Annas of Appied Statistics, 3: , 29. [25] P. Zhao, G. Rocha, and B. Yu. The composite absoute penaties famiy for grouped and hierarchica ariabe seection. Annas of Statistics, 37(6A): , 29. [26] L.W. Zhong and J.T. Kwok. Efficient sparse modeing with automatic feature grouping. In Internationa Conference on Machine Learning, 2. 9

Moreau-Yosida Regularization for Grouped Tree Structure Learning

Moreau-Yosida Regularization for Grouped Tree Structure Learning Moreau-Yosida Reguarization for Grouped Tree Structure Learning Jun Liu Computer Science and Engineering Arizona State University J.Liu@asu.edu Jieping Ye Computer Science and Engineering Arizona State

More information

Statistical Learning Theory: A Primer

Statistical Learning Theory: A Primer Internationa Journa of Computer Vision 38(), 9 3, 2000 c 2000 uwer Academic Pubishers. Manufactured in The Netherands. Statistica Learning Theory: A Primer THEODOROS EVGENIOU, MASSIMILIANO PONTIL AND TOMASO

More information

A Better Way to Pretrain Deep Boltzmann Machines

A Better Way to Pretrain Deep Boltzmann Machines A Better Way to Pretrain Deep Botzmann Machines Rusan Saakhutdino Department of Statistics and Computer Science Uniersity of Toronto rsaakhu@cs.toronto.edu Geoffrey Hinton Department of Computer Science

More information

Statistical Learning Theory: a Primer

Statistical Learning Theory: a Primer ??,??, 1 6 (??) c?? Kuwer Academic Pubishers, Boston. Manufactured in The Netherands. Statistica Learning Theory: a Primer THEODOROS EVGENIOU AND MASSIMILIANO PONTIL Center for Bioogica and Computationa

More information

MARKOV CHAINS AND MARKOV DECISION THEORY. Contents

MARKOV CHAINS AND MARKOV DECISION THEORY. Contents MARKOV CHAINS AND MARKOV DECISION THEORY ARINDRIMA DATTA Abstract. In this paper, we begin with a forma introduction to probabiity and expain the concept of random variabes and stochastic processes. After

More information

Stochastic Variational Inference with Gradient Linearization

Stochastic Variational Inference with Gradient Linearization Stochastic Variationa Inference with Gradient Linearization Suppementa Materia Tobias Pötz * Anne S Wannenwetsch Stefan Roth Department of Computer Science, TU Darmstadt Preface In this suppementa materia,

More information

Distributed average consensus: Beyond the realm of linearity

Distributed average consensus: Beyond the realm of linearity Distributed average consensus: Beyond the ream of inearity Usman A. Khan, Soummya Kar, and José M. F. Moura Department of Eectrica and Computer Engineering Carnegie Meon University 5 Forbes Ave, Pittsburgh,

More information

SUPPLEMENTARY MATERIAL TO INNOVATED SCALABLE EFFICIENT ESTIMATION IN ULTRA-LARGE GAUSSIAN GRAPHICAL MODELS

SUPPLEMENTARY MATERIAL TO INNOVATED SCALABLE EFFICIENT ESTIMATION IN ULTRA-LARGE GAUSSIAN GRAPHICAL MODELS ISEE 1 SUPPLEMENTARY MATERIAL TO INNOVATED SCALABLE EFFICIENT ESTIMATION IN ULTRA-LARGE GAUSSIAN GRAPHICAL MODELS By Yingying Fan and Jinchi Lv University of Southern Caifornia This Suppementary Materia

More information

Phase Change Equation of State for FSI Applications

Phase Change Equation of State for FSI Applications 15 th Internationa LS-DYNA Users Conference FSI / ALE Phase Change Equation of State for FSI Appications Mhamed Soui, Ramzi Messahe Lie Uniersity France Cyri Regan, Camie Ruiuc Ingeiance Technoogies, Agence

More information

Primal and dual active-set methods for convex quadratic programming

Primal and dual active-set methods for convex quadratic programming Math. Program., Ser. A 216) 159:469 58 DOI 1.17/s117-15-966-2 FULL LENGTH PAPER Prima and dua active-set methods for convex quadratic programming Anders Forsgren 1 Phiip E. Gi 2 Eizabeth Wong 2 Received:

More information

Appendix of the Paper The Role of No-Arbitrage on Forecasting: Lessons from a Parametric Term Structure Model

Appendix of the Paper The Role of No-Arbitrage on Forecasting: Lessons from a Parametric Term Structure Model Appendix of the Paper The Roe of No-Arbitrage on Forecasting: Lessons from a Parametric Term Structure Mode Caio Ameida cameida@fgv.br José Vicente jose.vaentim@bcb.gov.br June 008 1 Introduction In this

More information

FRST Multivariate Statistics. Multivariate Discriminant Analysis (MDA)

FRST Multivariate Statistics. Multivariate Discriminant Analysis (MDA) 1 FRST 531 -- Mutivariate Statistics Mutivariate Discriminant Anaysis (MDA) Purpose: 1. To predict which group (Y) an observation beongs to based on the characteristics of p predictor (X) variabes, using

More information

An Algorithm for Pruning Redundant Modules in Min-Max Modular Network

An Algorithm for Pruning Redundant Modules in Min-Max Modular Network An Agorithm for Pruning Redundant Modues in Min-Max Moduar Network Hui-Cheng Lian and Bao-Liang Lu Department of Computer Science and Engineering, Shanghai Jiao Tong University 1954 Hua Shan Rd., Shanghai

More information

arxiv: v1 [cs.lg] 31 Oct 2017

arxiv: v1 [cs.lg] 31 Oct 2017 ACCELERATED SPARSE SUBSPACE CLUSTERING Abofaz Hashemi and Haris Vikao Department of Eectrica and Computer Engineering, University of Texas at Austin, Austin, TX, USA arxiv:7.26v [cs.lg] 3 Oct 27 ABSTRACT

More information

A. Distribution of the test statistic

A. Distribution of the test statistic A. Distribution of the test statistic In the sequentia test, we first compute the test statistic from a mini-batch of size m. If a decision cannot be made with this statistic, we keep increasing the mini-batch

More information

SVM: Terminology 1(6) SVM: Terminology 2(6)

SVM: Terminology 1(6) SVM: Terminology 2(6) Andrew Kusiak Inteigent Systems Laboratory 39 Seamans Center he University of Iowa Iowa City, IA 54-57 SVM he maxima margin cassifier is simiar to the perceptron: It aso assumes that the data points are

More information

STABILITY OF A PARAMETRICALLY EXCITED DAMPED INVERTED PENDULUM 1. INTRODUCTION

STABILITY OF A PARAMETRICALLY EXCITED DAMPED INVERTED PENDULUM 1. INTRODUCTION Journa of Sound and Vibration (996) 98(5), 643 65 STABILITY OF A PARAMETRICALLY EXCITED DAMPED INVERTED PENDULUM G. ERDOS AND T. SINGH Department of Mechanica and Aerospace Engineering, SUNY at Buffao,

More information

Integrating Factor Methods as Exponential Integrators

Integrating Factor Methods as Exponential Integrators Integrating Factor Methods as Exponentia Integrators Borisav V. Minchev Department of Mathematica Science, NTNU, 7491 Trondheim, Norway Borko.Minchev@ii.uib.no Abstract. Recenty a ot of effort has been

More information

Optimal Blind Nonlinear Least-Squares Carrier Phase and Frequency Offset Estimation for Burst QAM Modulations

Optimal Blind Nonlinear Least-Squares Carrier Phase and Frequency Offset Estimation for Burst QAM Modulations Optima Bind Noninear Least-Squares Carrier Phase and Frequency Offset Estimation for Burst QAM Moduations Yan Wang Erchin Serpedin and Phiippe Cibat Dept. of Eectrica Engineering Texas A&M Uniersity Coege

More information

From Margins to Probabilities in Multiclass Learning Problems

From Margins to Probabilities in Multiclass Learning Problems From Margins to Probabiities in Muticass Learning Probems Andrea Passerini and Massimiiano Ponti 2 and Paoo Frasconi 3 Abstract. We study the probem of muticass cassification within the framework of error

More information

Appendix A: MATLAB commands for neural networks

Appendix A: MATLAB commands for neural networks Appendix A: MATLAB commands for neura networks 132 Appendix A: MATLAB commands for neura networks p=importdata('pn.xs'); t=importdata('tn.xs'); [pn,meanp,stdp,tn,meant,stdt]=prestd(p,t); for m=1:10 net=newff(minmax(pn),[m,1],{'tansig','purein'},'trainm');

More information

On the Goal Value of a Boolean Function

On the Goal Value of a Boolean Function On the Goa Vaue of a Booean Function Eric Bach Dept. of CS University of Wisconsin 1210 W. Dayton St. Madison, WI 53706 Lisa Heerstein Dept of CSE NYU Schoo of Engineering 2 Metrotech Center, 10th Foor

More information

NEW DEVELOPMENT OF OPTIMAL COMPUTING BUDGET ALLOCATION FOR DISCRETE EVENT SIMULATION

NEW DEVELOPMENT OF OPTIMAL COMPUTING BUDGET ALLOCATION FOR DISCRETE EVENT SIMULATION NEW DEVELOPMENT OF OPTIMAL COMPUTING BUDGET ALLOCATION FOR DISCRETE EVENT SIMULATION Hsiao-Chang Chen Dept. of Systems Engineering University of Pennsyvania Phiadephia, PA 904-635, U.S.A. Chun-Hung Chen

More information

International Journal of Mass Spectrometry

International Journal of Mass Spectrometry Internationa Journa of Mass Spectrometry 280 (2009) 179 183 Contents ists avaiabe at ScienceDirect Internationa Journa of Mass Spectrometry journa homepage: www.esevier.com/ocate/ijms Stark mixing by ion-rydberg

More information

arxiv: v2 [cs.lg] 4 Sep 2014

arxiv: v2 [cs.lg] 4 Sep 2014 Cassification with Sparse Overapping Groups Nikhi S. Rao Robert D. Nowak Department of Eectrica and Computer Engineering University of Wisconsin-Madison nrao2@wisc.edu nowak@ece.wisc.edu ariv:1402.4512v2

More information

DIGITAL FILTER DESIGN OF IIR FILTERS USING REAL VALUED GENETIC ALGORITHM

DIGITAL FILTER DESIGN OF IIR FILTERS USING REAL VALUED GENETIC ALGORITHM DIGITAL FILTER DESIGN OF IIR FILTERS USING REAL VALUED GENETIC ALGORITHM MIKAEL NILSSON, MATTIAS DAHL AND INGVAR CLAESSON Bekinge Institute of Technoogy Department of Teecommunications and Signa Processing

More information

Algorithms to solve massively under-defined systems of multivariate quadratic equations

Algorithms to solve massively under-defined systems of multivariate quadratic equations Agorithms to sove massivey under-defined systems of mutivariate quadratic equations Yasufumi Hashimoto Abstract It is we known that the probem to sove a set of randomy chosen mutivariate quadratic equations

More information

Melodic contour estimation with B-spline models using a MDL criterion

Melodic contour estimation with B-spline models using a MDL criterion Meodic contour estimation with B-spine modes using a MDL criterion Damien Loive, Ney Barbot, Oivier Boeffard IRISA / University of Rennes 1 - ENSSAT 6 rue de Kerampont, B.P. 80518, F-305 Lannion Cedex

More information

Inductive Bias: How to generalize on novel data. CS Inductive Bias 1

Inductive Bias: How to generalize on novel data. CS Inductive Bias 1 Inductive Bias: How to generaize on nove data CS 478 - Inductive Bias 1 Overfitting Noise vs. Exceptions CS 478 - Inductive Bias 2 Non-Linear Tasks Linear Regression wi not generaize we to the task beow

More information

Efficiently Generating Random Bits from Finite State Markov Chains

Efficiently Generating Random Bits from Finite State Markov Chains 1 Efficienty Generating Random Bits from Finite State Markov Chains Hongchao Zhou and Jehoshua Bruck, Feow, IEEE Abstract The probem of random number generation from an uncorreated random source (of unknown

More information

Available online at ScienceDirect. IFAC PapersOnLine 50-1 (2017)

Available online at   ScienceDirect. IFAC PapersOnLine 50-1 (2017) Avaiabe onine at www.sciencedirect.com ScienceDirect IFAC PapersOnLine 50-1 (2017 3412 3417 Stabiization of discrete-time switched inear systems: Lyapunov-Metzer inequaities versus S-procedure characterizations

More information

Lower Bounds for the Relative Greedy Algorithm for Approximating Steiner Trees

Lower Bounds for the Relative Greedy Algorithm for Approximating Steiner Trees This paper appeared in: Networks 47:2 (2006), -5 Lower Bounds for the Reative Greed Agorithm for Approimating Steiner Trees Stefan Hougard Stefan Kirchner Humbodt-Universität zu Berin Institut für Informatik

More information

Lecture Note 3: Stationary Iterative Methods

Lecture Note 3: Stationary Iterative Methods MATH 5330: Computationa Methods of Linear Agebra Lecture Note 3: Stationary Iterative Methods Xianyi Zeng Department of Mathematica Sciences, UTEP Stationary Iterative Methods The Gaussian eimination (or

More information

Uniprocessor Feasibility of Sporadic Tasks with Constrained Deadlines is Strongly conp-complete

Uniprocessor Feasibility of Sporadic Tasks with Constrained Deadlines is Strongly conp-complete Uniprocessor Feasibiity of Sporadic Tasks with Constrained Deadines is Strongy conp-compete Pontus Ekberg and Wang Yi Uppsaa University, Sweden Emai: {pontus.ekberg yi}@it.uu.se Abstract Deciding the feasibiity

More information

Maximum likelihood decoding of trellis codes in fading channels with no receiver CSI is a polynomial-complexity problem

Maximum likelihood decoding of trellis codes in fading channels with no receiver CSI is a polynomial-complexity problem 1 Maximum ikeihood decoding of treis codes in fading channes with no receiver CSI is a poynomia-compexity probem Chun-Hao Hsu and Achieas Anastasopouos Eectrica Engineering and Computer Science Department

More information

Iterative Decoding Performance Bounds for LDPC Codes on Noisy Channels

Iterative Decoding Performance Bounds for LDPC Codes on Noisy Channels Iterative Decoding Performance Bounds for LDPC Codes on Noisy Channes arxiv:cs/060700v1 [cs.it] 6 Ju 006 Chun-Hao Hsu and Achieas Anastasopouos Eectrica Engineering and Computer Science Department University

More information

ASummaryofGaussianProcesses Coryn A.L. Bailer-Jones

ASummaryofGaussianProcesses Coryn A.L. Bailer-Jones ASummaryofGaussianProcesses Coryn A.L. Baier-Jones Cavendish Laboratory University of Cambridge caj@mrao.cam.ac.uk Introduction A genera prediction probem can be posed as foows. We consider that the variabe

More information

FUSED MULTIPLE GRAPHICAL LASSO

FUSED MULTIPLE GRAPHICAL LASSO FUSED MULTIPLE GRAPHICAL LASSO SEN YANG, ZHAOSONG LU, XIAOTONG SHEN, PETER WONKA, JIEPING YE Abstract. In this paper, we consider the probem of estimating mutipe graphica modes simutaneousy using the fused

More information

Optimum Design Method of Viscous Dampers in Building Frames Using Calibration Model

Optimum Design Method of Viscous Dampers in Building Frames Using Calibration Model The 4 th Word Conference on Earthquake Engineering October -7, 8, Beijing, China Optimum Design Method of iscous Dampers in Buiding Frames sing Caibration Mode M. Yamakawa, Y. Nagano, Y. ee 3, K. etani

More information

Numerical Study on Subcooled Pool Boiling

Numerical Study on Subcooled Pool Boiling Progress in NUCLEAR SCIENCE and TECHNOLOGY, Vo., pp.15-19 (011) ARTICLE Numerica Study on Subcooed Poo Boiing Yasuo OSE * and Tomoaki KUNUGI Kyoto Uniersity, Yoshida, Sakyo, Kyoto, 606-8501, Japan This

More information

Approximated MLC shape matrix decomposition with interleaf collision constraint

Approximated MLC shape matrix decomposition with interleaf collision constraint Approximated MLC shape matrix decomposition with intereaf coision constraint Thomas Kainowski Antje Kiese Abstract Shape matrix decomposition is a subprobem in radiation therapy panning. A given fuence

More information

Explicit overall risk minimization transductive bound

Explicit overall risk minimization transductive bound 1 Expicit overa risk minimization transductive bound Sergio Decherchi, Paoo Gastado, Sandro Ridea, Rodofo Zunino Dept. of Biophysica and Eectronic Engineering (DIBE), Genoa University Via Opera Pia 11a,

More information

arxiv: v1 [math.co] 17 Dec 2018

arxiv: v1 [math.co] 17 Dec 2018 On the Extrema Maximum Agreement Subtree Probem arxiv:1812.06951v1 [math.o] 17 Dec 2018 Aexey Markin Department of omputer Science, Iowa State University, USA amarkin@iastate.edu Abstract Given two phyogenetic

More information

A proposed nonparametric mixture density estimation using B-spline functions

A proposed nonparametric mixture density estimation using B-spline functions A proposed nonparametric mixture density estimation using B-spine functions Atizez Hadrich a,b, Mourad Zribi a, Afif Masmoudi b a Laboratoire d Informatique Signa et Image de a Côte d Opae (LISIC-EA 4491),

More information

4 Separation of Variables

4 Separation of Variables 4 Separation of Variabes In this chapter we describe a cassica technique for constructing forma soutions to inear boundary vaue probems. The soution of three cassica (paraboic, hyperboic and eiptic) PDE

More information

Asynchronous Control for Coupled Markov Decision Systems

Asynchronous Control for Coupled Markov Decision Systems INFORMATION THEORY WORKSHOP (ITW) 22 Asynchronous Contro for Couped Marov Decision Systems Michae J. Neey University of Southern Caifornia Abstract This paper considers optima contro for a coection of

More information

Multilayer Kerceptron

Multilayer Kerceptron Mutiayer Kerceptron Zotán Szabó, András Lőrincz Department of Information Systems, Facuty of Informatics Eötvös Loránd University Pázmány Péter sétány 1/C H-1117, Budapest, Hungary e-mai: szzoi@csetehu,

More information

HILBERT? What is HILBERT? Matlab Implementation of Adaptive 2D BEM. Dirk Praetorius. Features of HILBERT

HILBERT? What is HILBERT? Matlab Implementation of Adaptive 2D BEM. Dirk Praetorius. Features of HILBERT Söerhaus-Workshop 2009 October 16, 2009 What is HILBERT? HILBERT Matab Impementation of Adaptive 2D BEM joint work with M. Aurada, M. Ebner, S. Ferraz-Leite, P. Godenits, M. Karkuik, M. Mayr Hibert Is

More information

Determining The Degree of Generalization Using An Incremental Learning Algorithm

Determining The Degree of Generalization Using An Incremental Learning Algorithm Determining The Degree of Generaization Using An Incrementa Learning Agorithm Pabo Zegers Facutad de Ingeniería, Universidad de os Andes San Caros de Apoquindo 22, Las Condes, Santiago, Chie pzegers@uandes.c

More information

Sequential Decoding of Polar Codes with Arbitrary Binary Kernel

Sequential Decoding of Polar Codes with Arbitrary Binary Kernel Sequentia Decoding of Poar Codes with Arbitrary Binary Kerne Vera Miosavskaya, Peter Trifonov Saint-Petersburg State Poytechnic University Emai: veram,petert}@dcn.icc.spbstu.ru Abstract The probem of efficient

More information

An analysis of the dynamics of a launcher-missile system on a moveable base

An analysis of the dynamics of a launcher-missile system on a moveable base BULLETIN OF THE POLISH ACADEMY OF SCIENCES TECHNICAL SCIENCES, Vo. 58, No. 4, 010 DOI: 10.478/10175-010-0068-5 VARIA An anaysis of the dynamics of a auncher-missie system on a moeabe base Z. DZIOPA, I.

More information

Introduction to Simulation - Lecture 13. Convergence of Multistep Methods. Jacob White. Thanks to Deepak Ramaswamy, Michal Rewienski, and Karen Veroy

Introduction to Simulation - Lecture 13. Convergence of Multistep Methods. Jacob White. Thanks to Deepak Ramaswamy, Michal Rewienski, and Karen Veroy Introduction to Simuation - Lecture 13 Convergence of Mutistep Methods Jacob White Thans to Deepa Ramaswamy, Micha Rewiensi, and Karen Veroy Outine Sma Timestep issues for Mutistep Methods Loca truncation

More information

BP neural network-based sports performance prediction model applied research

BP neural network-based sports performance prediction model applied research Avaiabe onine www.jocpr.com Journa of Chemica and Pharmaceutica Research, 204, 6(7:93-936 Research Artice ISSN : 0975-7384 CODEN(USA : JCPRC5 BP neura networ-based sports performance prediction mode appied

More information

CS229 Lecture notes. Andrew Ng

CS229 Lecture notes. Andrew Ng CS229 Lecture notes Andrew Ng Part IX The EM agorithm In the previous set of notes, we taked about the EM agorithm as appied to fitting a mixture of Gaussians. In this set of notes, we give a broader view

More information

Akaike Information Criterion for ANOVA Model with a Simple Order Restriction

Akaike Information Criterion for ANOVA Model with a Simple Order Restriction Akaike Information Criterion for ANOVA Mode with a Simpe Order Restriction Yu Inatsu * Department of Mathematics, Graduate Schoo of Science, Hiroshima University ABSTRACT In this paper, we consider Akaike

More information

Construction of Supersaturated Design with Large Number of Factors by the Complementary Design Method

Construction of Supersaturated Design with Large Number of Factors by the Complementary Design Method Acta Mathematicae Appicatae Sinica, Engish Series Vo. 29, No. 2 (2013) 253 262 DOI: 10.1007/s10255-013-0214-6 http://www.appmath.com.cn & www.springerlink.com Acta Mathema cae Appicatae Sinica, Engish

More information

Available online at ScienceDirect. Procedia Computer Science 96 (2016 )

Available online at  ScienceDirect. Procedia Computer Science 96 (2016 ) Avaiabe onine at www.sciencedirect.com ScienceDirect Procedia Computer Science 96 (206 92 99 20th Internationa Conference on Knowedge Based and Inteigent Information and Engineering Systems Connected categorica

More information

ORTHOGONAL MULTI-WAVELETS FROM MATRIX FACTORIZATION

ORTHOGONAL MULTI-WAVELETS FROM MATRIX FACTORIZATION J. Korean Math. Soc. 46 2009, No. 2, pp. 281 294 ORHOGONAL MLI-WAVELES FROM MARIX FACORIZAION Hongying Xiao Abstract. Accuracy of the scaing function is very crucia in waveet theory, or correspondingy,

More information

A Ridgelet Kernel Regression Model using Genetic Algorithm

A Ridgelet Kernel Regression Model using Genetic Algorithm A Ridgeet Kerne Regression Mode using Genetic Agorithm Shuyuan Yang, Min Wang, Licheng Jiao * Institute of Inteigence Information Processing, Department of Eectrica Engineering Xidian University Xi an,

More information

BALANCING REGULAR MATRIX PENCILS

BALANCING REGULAR MATRIX PENCILS BALANCING REGULAR MATRIX PENCILS DAMIEN LEMONNIER AND PAUL VAN DOOREN Abstract. In this paper we present a new diagona baancing technique for reguar matrix pencis λb A, which aims at reducing the sensitivity

More information

In-plane shear stiffness of bare steel deck through shell finite element models. G. Bian, B.W. Schafer. June 2017

In-plane shear stiffness of bare steel deck through shell finite element models. G. Bian, B.W. Schafer. June 2017 In-pane shear stiffness of bare stee deck through she finite eement modes G. Bian, B.W. Schafer June 7 COLD-FORMED STEEL RESEARCH CONSORTIUM REPORT SERIES CFSRC R-7- SDII Stee Diaphragm Innovation Initiative

More information

A Brief Introduction to Markov Chains and Hidden Markov Models

A Brief Introduction to Markov Chains and Hidden Markov Models A Brief Introduction to Markov Chains and Hidden Markov Modes Aen B MacKenzie Notes for December 1, 3, &8, 2015 Discrete-Time Markov Chains You may reca that when we first introduced random processes,

More information

Semidefinite relaxation and Branch-and-Bound Algorithm for LPECs

Semidefinite relaxation and Branch-and-Bound Algorithm for LPECs Semidefinite reaxation and Branch-and-Bound Agorithm for LPECs Marcia H. C. Fampa Universidade Federa do Rio de Janeiro Instituto de Matemática e COPPE. Caixa Posta 68530 Rio de Janeiro RJ 21941-590 Brasi

More information

SVM-based Supervised and Unsupervised Classification Schemes

SVM-based Supervised and Unsupervised Classification Schemes SVM-based Supervised and Unsupervised Cassification Schemes LUMINITA STATE University of Pitesti Facuty of Mathematics and Computer Science 1 Targu din Vae St., Pitesti 110040 ROMANIA state@cicknet.ro

More information

An explicit Jordan Decomposition of Companion matrices

An explicit Jordan Decomposition of Companion matrices An expicit Jordan Decomposition of Companion matrices Fermín S V Bazán Departamento de Matemática CFM UFSC 88040-900 Forianópois SC E-mai: fermin@mtmufscbr S Gratton CERFACS 42 Av Gaspard Coriois 31057

More information

Worst Case Analysis of the Analog Circuits

Worst Case Analysis of the Analog Circuits Proceedings of the 11th WSEAS Internationa Conference on CIRCUITS, Agios Nikoaos, Crete Isand, Greece, Juy 3-5, 7 9 Worst Case Anaysis of the Anaog Circuits ELENA NICULESCU*, DORINA-MIOARA PURCARU* and

More information

Convergence Property of the Iri-Imai Algorithm for Some Smooth Convex Programming Problems

Convergence Property of the Iri-Imai Algorithm for Some Smooth Convex Programming Problems Convergence Property of the Iri-Imai Agorithm for Some Smooth Convex Programming Probems S. Zhang Communicated by Z.Q. Luo Assistant Professor, Department of Econometrics, University of Groningen, Groningen,

More information

Approximated MLC shape matrix decomposition with interleaf collision constraint

Approximated MLC shape matrix decomposition with interleaf collision constraint Agorithmic Operations Research Vo.4 (29) 49 57 Approximated MLC shape matrix decomposition with intereaf coision constraint Antje Kiese and Thomas Kainowski Institut für Mathematik, Universität Rostock,

More information

Learning Fully Observed Undirected Graphical Models

Learning Fully Observed Undirected Graphical Models Learning Fuy Observed Undirected Graphica Modes Sides Credit: Matt Gormey (2016) Kayhan Batmangheich 1 Machine Learning The data inspires the structures we want to predict Inference finds {best structure,

More information

A Comparison Study of the Test for Right Censored and Grouped Data

A Comparison Study of the Test for Right Censored and Grouped Data Communications for Statistica Appications and Methods 2015, Vo. 22, No. 4, 313 320 DOI: http://dx.doi.org/10.5351/csam.2015.22.4.313 Print ISSN 2287-7843 / Onine ISSN 2383-4757 A Comparison Study of the

More information

TUNING PARAMETER SELECTION FOR PENALIZED LIKELIHOOD ESTIMATION OF GAUSSIAN GRAPHICAL MODEL

TUNING PARAMETER SELECTION FOR PENALIZED LIKELIHOOD ESTIMATION OF GAUSSIAN GRAPHICAL MODEL Statistica Sinica 22 (2012), 1123-1146 doi:http://dx.doi.org/10.5705/ss.2009.210 TUNING PARAMETER SELECTION FOR PENALIZED LIKELIHOOD ESTIMATION OF GAUSSIAN GRAPHICAL MODEL Xin Gao, Danie Q. Pu, Yuehua

More information

(This is a sample cover image for this issue. The actual cover is not yet available at this time.)

(This is a sample cover image for this issue. The actual cover is not yet available at this time.) (This is a sampe cover image for this issue The actua cover is not yet avaiabe at this time) This artice appeared in a journa pubished by Esevier The attached copy is furnished to the author for interna

More information

Problem set 6 The Perron Frobenius theorem.

Problem set 6 The Perron Frobenius theorem. Probem set 6 The Perron Frobenius theorem. Math 22a4 Oct 2 204, Due Oct.28 In a future probem set I want to discuss some criteria which aow us to concude that that the ground state of a sef-adjoint operator

More information

On a geometrical approach in contact mechanics

On a geometrical approach in contact mechanics Institut für Mechanik On a geometrica approach in contact mechanics Aexander Konyukhov, Kar Schweizerhof Universität Karsruhe, Institut für Mechanik Institut für Mechanik Kaiserstr. 12, Geb. 20.30 76128

More information

Numerical Simulation for Optimizing Temperature Gradients during Single Crystal Casting Process

Numerical Simulation for Optimizing Temperature Gradients during Single Crystal Casting Process ISIJ Internationa Vo 54 (2014) No 2 pp 254 258 Numerica Simuation for Optimizing Temperature Gradients during Singe Crysta Casting Process Aeksandr Aeksandrovich INOZEMTSEV 1) Aeksandra Sergeevna DUBROVSKAYA

More information

Coupling of LWR and phase transition models at boundary

Coupling of LWR and phase transition models at boundary Couping of LW and phase transition modes at boundary Mauro Garaveo Dipartimento di Matematica e Appicazioni, Università di Miano Bicocca, via. Cozzi 53, 20125 Miano Itay. Benedetto Piccoi Department of

More information

Soft Clustering on Graphs

Soft Clustering on Graphs Soft Custering on Graphs Kai Yu 1, Shipeng Yu 2, Voker Tresp 1 1 Siemens AG, Corporate Technoogy 2 Institute for Computer Science, University of Munich kai.yu@siemens.com, voker.tresp@siemens.com spyu@dbs.informatik.uni-muenchen.de

More information

A Novel Learning Method for Elman Neural Network Using Local Search

A Novel Learning Method for Elman Neural Network Using Local Search Neura Information Processing Letters and Reviews Vo. 11, No. 8, August 2007 LETTER A Nove Learning Method for Eman Neura Networ Using Loca Search Facuty of Engineering, Toyama University, Gofuu 3190 Toyama

More information

Interactive Fuzzy Programming for Two-level Nonlinear Integer Programming Problems through Genetic Algorithms

Interactive Fuzzy Programming for Two-level Nonlinear Integer Programming Problems through Genetic Algorithms Md. Abu Kaam Azad et a./asia Paciic Management Review (5) (), 7-77 Interactive Fuzzy Programming or Two-eve Noninear Integer Programming Probems through Genetic Agorithms Abstract Md. Abu Kaam Azad a,*,

More information

Published in: Proceedings of the Twenty Second Nordic Seminar on Computational Mechanics

Published in: Proceedings of the Twenty Second Nordic Seminar on Computational Mechanics Aaborg Universitet An Efficient Formuation of the Easto-pastic Constitutive Matrix on Yied Surface Corners Causen, Johan Christian; Andersen, Lars Vabbersgaard; Damkide, Lars Pubished in: Proceedings of

More information

First-Order Corrections to Gutzwiller s Trace Formula for Systems with Discrete Symmetries

First-Order Corrections to Gutzwiller s Trace Formula for Systems with Discrete Symmetries c 26 Noninear Phenomena in Compex Systems First-Order Corrections to Gutzwier s Trace Formua for Systems with Discrete Symmetries Hoger Cartarius, Jörg Main, and Günter Wunner Institut für Theoretische

More information

Power Control and Transmission Scheduling for Network Utility Maximization in Wireless Networks

Power Control and Transmission Scheduling for Network Utility Maximization in Wireless Networks ower Contro and Transmission Scheduing for Network Utiity Maximization in Wireess Networks Min Cao, Vivek Raghunathan, Stephen Hany, Vinod Sharma and. R. Kumar Abstract We consider a joint power contro

More information

Input-to-state stability for a class of Lurie systems

Input-to-state stability for a class of Lurie systems Automatica 38 (2002) 945 949 www.esevier.com/ocate/automatica Brief Paper Input-to-state stabiity for a cass of Lurie systems Murat Arcak a;, Andrew Tee b a Department of Eectrica, Computer and Systems

More information

STA 216 Project: Spline Approach to Discrete Survival Analysis

STA 216 Project: Spline Approach to Discrete Survival Analysis : Spine Approach to Discrete Surviva Anaysis November 4, 005 1 Introduction Athough continuous surviva anaysis differs much from the discrete surviva anaysis, there is certain ink between the two modeing

More information

Nonlinear Gaussian Filtering via Radial Basis Function Approximation

Nonlinear Gaussian Filtering via Radial Basis Function Approximation 51st IEEE Conference on Decision and Contro December 10-13 01 Maui Hawaii USA Noninear Gaussian Fitering via Radia Basis Function Approximation Huazhen Fang Jia Wang and Raymond A de Caafon Abstract This

More information

Another Look at Linear Programming for Feature Selection via Methods of Regularization 1

Another Look at Linear Programming for Feature Selection via Methods of Regularization 1 Another Look at Linear Programming for Feature Seection via Methods of Reguarization Yonggang Yao, The Ohio State University Yoonkyung Lee, The Ohio State University Technica Report No. 800 November, 2007

More information

A Solution to the 4-bit Parity Problem with a Single Quaternary Neuron

A Solution to the 4-bit Parity Problem with a Single Quaternary Neuron Neura Information Processing - Letters and Reviews Vo. 5, No. 2, November 2004 LETTER A Soution to the 4-bit Parity Probem with a Singe Quaternary Neuron Tohru Nitta Nationa Institute of Advanced Industria

More information

Proceedings of the 2012 Winter Simulation Conference C. Laroque, J. Himmelspach, R. Pasupathy, O. Rose, and A. M. Uhrmacher, eds.

Proceedings of the 2012 Winter Simulation Conference C. Laroque, J. Himmelspach, R. Pasupathy, O. Rose, and A. M. Uhrmacher, eds. Proceedings of the 2012 Winter Simuation Conference C. aroque, J. Himmespach, R. Pasupathy, O. Rose, and A. M. Uhrmacher, eds. TIGHT BOUNDS FOR AMERICAN OPTIONS VIA MUTIEVE MONTE CARO Denis Beomestny Duisburg-Essen

More information

Extended SMART Algorithms for Non-Negative Matrix Factorization

Extended SMART Algorithms for Non-Negative Matrix Factorization Extended SMART Agorithms for Non-Negative Matrix Factorization Andrzej CICHOCKI 1, Shun-ichi AMARI 2 Rafa ZDUNEK 1, Rau KOMPASS 1, Gen HORI 1 and Zhaohui HE 1 Invited Paper 1 Laboratory for Advanced Brain

More information

arxiv: v2 [stat.ml] 19 Oct 2016

arxiv: v2 [stat.ml] 19 Oct 2016 Sparse Quadratic Discriminant Anaysis and Community Bayes arxiv:1407.4543v2 [stat.ml] 19 Oct 2016 Ya Le Department of Statistics Stanford University ye@stanford.edu Abstract Trevor Hastie Department of

More information

Mixed Volume Computation, A Revisit

Mixed Volume Computation, A Revisit Mixed Voume Computation, A Revisit Tsung-Lin Lee, Tien-Yien Li October 31, 2007 Abstract The superiority of the dynamic enumeration of a mixed ces suggested by T Mizutani et a for the mixed voume computation

More information

Research Article Optimal Control of Probabilistic Logic Networks and Its Application to Real-Time Pricing of Electricity

Research Article Optimal Control of Probabilistic Logic Networks and Its Application to Real-Time Pricing of Electricity Hindawi Pubishing Corporation Mathematica Probems in Engineering Voume 05, Artice ID 950, 0 pages http://dxdoiorg/055/05/950 Research Artice Optima Contro of Probabiistic Logic Networks and Its Appication

More information

A SIMPLIFIED DESIGN OF MULTIDIMENSIONAL TRANSFER FUNCTION MODELS

A SIMPLIFIED DESIGN OF MULTIDIMENSIONAL TRANSFER FUNCTION MODELS A SIPLIFIED DESIGN OF ULTIDIENSIONAL TRANSFER FUNCTION ODELS Stefan Petrausch, Rudof Rabenstein utimedia Communications and Signa Procesg, University of Erangen-Nuremberg, Cauerstr. 7, 958 Erangen, GERANY

More information

Numerical solution of one dimensional contaminant transport equation with variable coefficient (temporal) by using Haar wavelet

Numerical solution of one dimensional contaminant transport equation with variable coefficient (temporal) by using Haar wavelet Goba Journa of Pure and Appied Mathematics. ISSN 973-1768 Voume 1, Number (16), pp. 183-19 Research India Pubications http://www.ripubication.com Numerica soution of one dimensiona contaminant transport

More information

Some Measures for Asymmetry of Distributions

Some Measures for Asymmetry of Distributions Some Measures for Asymmetry of Distributions Georgi N. Boshnakov First version: 31 January 2006 Research Report No. 5, 2006, Probabiity and Statistics Group Schoo of Mathematics, The University of Manchester

More information

XSAT of linear CNF formulas

XSAT of linear CNF formulas XSAT of inear CN formuas Bernd R. Schuh Dr. Bernd Schuh, D-50968 Kön, Germany; bernd.schuh@netcoogne.de eywords: compexity, XSAT, exact inear formua, -reguarity, -uniformity, NPcompeteness Abstract. Open

More information

Minimum Enclosing Circle of a Set of Fixed Points and a Mobile Point

Minimum Enclosing Circle of a Set of Fixed Points and a Mobile Point Minimum Encosing Circe of a Set of Fixed Points and a Mobie Point Aritra Banik 1, Bhaswar B. Bhattacharya 2, and Sandip Das 1 1 Advanced Computing and Microeectronics Unit, Indian Statistica Institute,

More information

Probabilistic Graphical Models

Probabilistic Graphical Models Schoo of Computer Science Probabiistic Graphica Modes Gaussian graphica modes and Ising modes: modeing networks Eric Xing Lecture 0, February 0, 07 Reading: See cass website Eric Xing @ CMU, 005-07 Network

More information

Haar Decomposition and Reconstruction Algorithms

Haar Decomposition and Reconstruction Algorithms Jim Lambers MAT 773 Fa Semester 018-19 Lecture 15 and 16 Notes These notes correspond to Sections 4.3 and 4.4 in the text. Haar Decomposition and Reconstruction Agorithms Decomposition Suppose we approximate

More information

Theory of Generalized k-difference Operator and Its Application in Number Theory

Theory of Generalized k-difference Operator and Its Application in Number Theory Internationa Journa of Mathematica Anaysis Vo. 9, 2015, no. 19, 955-964 HIKARI Ltd, www.m-hiari.com http://dx.doi.org/10.12988/ijma.2015.5389 Theory of Generaized -Difference Operator and Its Appication

More information