Automatic Name-Face Alignment to Enable Cross-Media News Retrieval

Size: px
Start display at page:

Download "Automatic Name-Face Alignment to Enable Cross-Media News Retrieval"

Transcription

1 Proceedings of the Twenty-Third International Joint Conference on Artificial Intelligence Autoatic ae-face Alignent to Enable Cross-Media ews Retrieval Yueie Zhang*, Wei Wu*, Yang Li*, Cheng Jin*, Xiangyang Xue*, Jianping Fan + *School of Coputer Science, Shanghai Key Laboratory of Intelligent Inforation Processing, Fudan University, Shanghai, China + Departent of Coputer Science, The University of orth Carolina at Charlotte, USA *{yzhang, , , c, xyxue}@fudan.edu.cn, + fan@uncc.edu Abstract A new algorith is developed in this paper to support autoatic nae-face alignent for achieving ore accurate cross-edia news retrieval. We focus on extracting valuable inforation fro large aounts of news iages and their captions, where ulti-level iage-caption pairs are constructed for characterizing both significant naes with higher salience and their cohesion with huan faces extracted fro news iages. To reedy the issue of lacking enough related inforation for rare nae, Web ining is introduced to acquire the extra ultiodal inforation. We also ephasize on an optiization echanis by our Iproved Self-Adaptive Siulated Annealing Genetic Algorith to verify the feasibility of alignent cobinations. Our experients have obtained very positive results. 1 Introduction With the explosive growth of ultiodal news available both online and offline, how to integrate ultiodal inforation sources to achieve ore accurate cross-edia news retrieval becoes an iportant research focus [Datta et al., 2008]. Usually ultiodal news is exhibited with the for of captioned news iages, which ostly describe stories about people [Liu et al., 2008]. Thus enabling autoatic nae-face alignent has becoe a critical issue for supporting cross-edia news retrieval. However, because of the seantic gap [Fan et al., 2012], there ay exist huge uncertainty on the correspondence relationships aong naes (in captions) and faces (in iages) [Deschacht et al., 2007], [Yang et al., 2009], e.g., the relationship between naes and faces in a news is any-to-any rather than one-to-one. To achieve autoatic nae-face alignent, three inter-related issues should be addressed siultaneously: 1) ultiodal analysis of cross-edia news to identify better correspondences between naes and faces; 2) discovery of the issing inforation for rare nae; and 3) ultiodal optiization to achieve ore effective alignents for all possible nae-face pairs. To address the first issue, it is very iportant to develop robust algoriths that are able to achieve ore accurate nae-face alignent. To address the second issue, it is very interesting to leverage large-scale Web news for issing inforation prediction. To address the third issue, it is critical to develop a new optiization algorith with high accuracy rate but low coputational cost. Based on these observations above, a novel schee is developed in this paper for facilitating autoatic nae-face alignent to enable ore effective cross-edia news retrieval. Our schee significantly differs fro other earlier work in: a) Salient naes and cohesive faces are extracted fro cross-edia news. Such naes ight have higher correspondence possibility with the faces in news iages, and such faces ight have higher coherence with the salient naes. b) Web news is integrated for discovering issing inforation for rare nae. c) An efficient easureent and optiization echanis is established to verify the feasibility of our autoatic alignent algorith. d) A new algorith is developed to achieve ore precise characterization of the correlations between naes and faces. Such cross-edia alignent is treated as a proble of bi-edia seantic apping, and odeled as a correlation distribution over seantic representations of naes and faces. Our experients on a large nuber of public data fro Yahoo! ews have obtained very positive results. 2 Related Work Alignent of naes and faces in cross-edia news is not a novel task [Satoh et al., 1997&1999], [Berg et al., 2004&2005]. Earlier research considered acquiring the relevant faces based on the original text query over iage captions, and ranking or filtering the returned iages by a face detector. However, the irrelevant or incorrect results ay be produced by utilizing the siple atching between a query nae and captions. Most face recognition approaches are only applied to the controlled setting and liited data collections [Mensink et al., 2008], [Bozorgtabar et al., 2011]. Thus the increasing research interest focuses on cobining textual and visual inforation to support precise alignent. In recent years, there is soe related research work by using textual inforation that accopanies the iage. Yang et al. [2004] proposed an approach for finding the specific persons in broadcast news videos by exploring various clues such as naes in the transcript, face, anchor scenes and the tiing pattern between naes and people. Everingha et al. 2768

2 [2006], [2009] investigated the proble of autoatically labeling appearances of the naes in TV or fil, and deonstrated that high precision can be achieved by cobining ultiple sources of inforation. Ozkan et al. [2006] proposed a graph-based ethod to retrieve the correct faces of a queried person using both text and visual appearances. Le et al. [2007], [2008] introduced an unsupervised ethod for face annotation by ining Web, which aied to retrieve relevant faces of one person by learning the visual consistency aong results retrieved fro text-correlation-based search engines. It s necessary to ention the ost interesting research work developed by Berg et al. [2004], [2005], [2007], Guillauin et al. [2008], [2012], and Pha et al. [2010], [2011]. Berg et al. proposed a ethod for association of naes to faces using ore realistic dataset. Guillauin et al. considered two scenarios of naing people in databases of news photos with captions, that is, finding faces of a single query person and assigning naes to all faces. Pha et al. reported their experients on aligning naes and faces as found in the iages and captions of online news websites. Unfortunately, all these existing approaches have not provided good solutions for the following iportant issues: (1) Analysis and Measure for ae Salience and ae-face Cohesion in Deep Level Most existing alignent ethods focus on exploiting all the inter-linkage inforation between each nae and each face in a sae captioned news iage, which leads to a serious cobinatorial proble for possible nae-face pairs. Without sufficient analysis of nae salience and easure for nae-face cohesion, soe naes and faces ay for the noisy inforation and cause the insignificant nae-face atching udgeent. (2) Discovering Extra Multiodal Inforation for Rare ae Most existing work concerns finding faces of a specific nae underlying such a precondition that the relevant face set consists of a large group of highly siilar faces for the original nae query. With the increasing growth of Web news, it appears that such Web inforation can itigate the lack of relevant faces for rare nae. According to our best knowledge, no existing research has ade full use of Web news to achieve ore accurate alignent for rare nae. (3) Multiodal Alignent Optiization for ae-face Pairs Finding the best atching between naes and faces for all the pairs is very difficult and coplex, and ay becoe a P-hard proble. Probability-statistical odels can be used for coputing the alignent likelihood, but have the serious proble of getting stuck in a local axiu. Fro the view of cobinatorial optiization, it s a very significant way to set a global obective function, the local constraint conditions and an integer prograing odel. 3 ae Salience Identification In a news caption, not all naes are equally iportant. Thus it s necessary to evaluate every nae in the sae caption and udge which naes ay possibly co-occur with the faces in the original iage-caption pair. Here, salience is defined as a easureent for the iportant degree of each nae in a caption. A pattern for coputing such a score is constructed based on the in-depth ulti-level analysis of caption. In the initial nae list, the naes are enuerated without any ranking. Since the syntactic structure is indicative of various inforation distribution in a caption, the relative rank of naes can be confired by eans of the roles of naes and the relationships aong naes in the syntactic parse tree. (1) Syntactic Parse Tree Depth (SPTD) This factor indicates the iniu depth for all the naes of a certain nae clustering in the parse tree for a caption. If a nae has a shallower depth, it will have the higher iportance. Given a caption with naes, each nae is associated with its corresponding nae clustering C i and C i represents the th nae in C i. The depth value for each C i can be defined as: C Ci SC SPT Depth Ci i SPTD in 1 _ (1) where SC(C i ) denotes the size of C i, i.e., the nuber of naes with the co-reference relationship contained in C i, which can also be understood as the occurrence frequency of the sae naes with different fors in the given caption; and SPT_Depth(C i) is the depth of C i in the parse tree. (2) Syntactic Parse Tree Traversal Order (SPTTO) This factor indicates the iniu breadth-first traversal order for all the naes of a certain nae clustering in the parse tree for a caption. It s proposed based on such a fact that aong the nodes in the sae level, a node with the higher priority of the traversal order eans its higher relative iportance. The traversal order value for each C i can be defined as: C Ci SC SPT _ BFT OrderCi i SPTTO where SPT_BFT-Order(C i) is the breadth-first traversal order of C i in the syntactic parse tree. In ost cases the higher frequency for a specific nae in a caption indicates its higher iportance. It s very necessary to take such frequency into account as a key reference factor. Given different naes in a caption, in which each nae corresponds to a nae clustering C i, i=1,,, the initial rank value, Relative Salience (RS), can be coputed as: in (2) 1 SC Ci 1SPTDC SPTDCi RS Ci (3) 1 SCC 1 1SPTDC 1 SPTTOC SPTTOCi, i1 RS Ci SPTTOC where α, β and γ are the relative iportance coefficients, α+β+γ=1. Maybe there is an extree situation that only one nae is involved in the caption, it s insignificant to evaluate the relative salience for such a unique nae, RS(C 1 )=1. It s rather crude to assue that every nae in the caption appears in the iage. Thus it s very significant to ake the further refineent to filter the naes with the extreely lower possibility to be pictured as the detected faces. A visual survey for nae-face pairs with the association linkage learned that the nuber of available faces is generally less than or equal to the nuber of detected naes in a pair. We ake the siplifying assuption that all the naes with the initial rank values below a certain threshold ay not be pictured as the selected faces. Given F faces and naes (F<=) in the initial rank list for a pair, F selected faces (F <=F) are retained after face filtering and naes are refined to naes (F <= <=) according to the visual exhibition inforation. Under such refineent, ultiple naes with the relatively higher iportance are chosen to interpret the selected faces in the iage, i.e., nae denoising. 2769

3 4 ae-face Cohesion Measure The general assuption for nae-face alignent is that the appearances of the faces for a person/nae are ore siilar than those for different persons/naes. Thus it s very useful to establish a cohesion easureent to evaluate the intrinsic association degree aong all the faces in the whole database. Firstly, the siilarity between two faces is evaluated based on the coon neighbors aong k-nearest neighbors for every face. Given two faces F i and F in the local face set FS_ related to a real nae, if F i is close to F and they are both close to FS_, F i and F are close with higher confidence for the utual density since their siilarity is deterined based on all the faces in FS_. Thus the siilarity between a pair of faces can be defined as: KSFi, FS _, k KSF, FS _, k SiFi, F (4) k where KS(F i, FS_, k) and KS(F, FS_, k) denote the k-nearest neighbors when F i or F positions in FS_ ; and k is a dynaically varied value and changes according to the size of the local face set related to the current nae. Secondly, the Local Density Score (LDS) is utilized to easure the density for each face in the local face set. The larger the density is, the ore relevant the face and the nae associated with the face set are. The LDS for a face F i can be described as the average siilarity between this face and its k-nearest faces in the sae face set. The higher LDS indicates the higher connectivity between F i and its k-nearest faces, that is, F i is ore relevant with associated with FS_. F KSF FS k SiFi F i LDSFi FS k, _,,, _, (5) k Thirdly, the Local Cohesion Degree (LCD) is especially proposed to easure the utual density aong all the faces in the face set related to a particular nae under the current global alignent anner, and defined as: FS, k LDSFi, FS _ k LCD _ FiFS _, (6) If a face has the higher LDS value in the related face set, this face will have the higher density with the other faces in the sae face set and then will be ore relevant with the nae corresponding to the face set. Therefore, as the su of the LDS values for all the faces in a local face set, LCD can be well representative of the utual density aong all the faces in this face set. The larger the LCD is, the higher the utual density aong all the faces in the face set will be. The higher LCD can be regarded as a good indication of the higher local cohesion for the face set, that is, the face set is ore relevant with the assigned nae. Under an arbitrary global alignent anner, the su of the LCD values for all the face sets can reflect the whole global cohesion for the coplete face set related to all the naes. Thus the larger su of the LCD values for the face sets ebodies the higher global cohesion for the coplete face set, and then further expresses the higher global relevance between the whole face set and the nae set under the current global alignent anner. 5 Web Mining for Rare ae The nae-face alignent can also be regarded as a restricted face naing or retrieval proble. The general assuption is that the nuber of faces corresponding to the query nae is relatively large. However, soe rare naes ay lead to such a fact that the nuber of their faces is very sall. The rare nae here does not ean the real single nae, but refers to the rare person that occurs only several ties or even one tie in the whole database. Thus for the alignent evaluation of rare nae, it s necessary to establish a discovery echanis to suppleent ore available ultiodal inforation. As the vast knowledge base, Web has becoe the best resource to create a reference annotated face iage set for rare nae. Face iages about a specific person can be autoatically retrieved by using Web iage search engines. Based on Web ining, a certain nuber of relevant face iages and their annotations can be discovered siultaneously, and becoe the valuable enrichent for rare nae. All the nae expression fors for a rare nae are collected in a rare nae clustering. Each for in the clustering is taken as an original nae query and subitted to Web iage search engines. Thus a list of relevant face iages and their annotations can both be acquired. Each returned face iage will be processed through the face detection and filtering, and the available and eaningful face iages are kept. Given a rare nae ae 0, all of its face iages can be ranked according to the salience value of ae 0 in each face iage. The Top-R face iages are aintained as the copleentary face inforation to ae 0. R is a generalized threshold for the copleentary face iage selection, and does not have a fixed setting. Because there ay be different nubers of face iages returned for different rare naes, R eans a liited proportion of the face iages preferred, and varies with the size of the returned face iage set. 6 ae-face Alignent via ISSAGA We consider such an alignent task as an optiization proble for searching the optial nae-face atching in each pair. The Standard Genetic Algorith (SGA) can obtain the global optial search using genetic operations, but there are soe disadvantages such as preaturity and the poor search power for local optial solutions [Wang et al., 2005]. However, the Siulated Annealing Algorith (SAA) has the stronger ability for local optiization [Andresen et al., 2008]. Thus the Iproved Self-Adaptive Siulated Annealing Genetic Algorith (ISSAGA), which integrates the advantages of SGA and SAA and a self-adaptive probability adustent strategy, is introduced to ake a better solution. 6.1 Multiodal ae-face Alignent Suppose P iage-caption pairs include F real faces (i.e., ot ull Face) and real naes (i.e., ot ull ae), PS denotes the Iage-Caption Pair Set for P pairs; FS denotes the Face Set for F faces; S denotes the ae Set for naes; W_FP i denotes whether the face F belongs to the pair P i, P i PS, F FS, i=1,, PS, =1,, FS, which is equal to: 1, if the face F is inthe pair Pi W _ FPi 0, otherwise W_P ik denotes whether the nae k belongs to P i, P i S, i=1,, PS, k=1,, S, which is equal to: k PS, 2770

4 1, if the nae k is in the pair Pi W _ Pik 0, otherwise FP i denotes the face set for P i, FP i ={F W_FP i=1, F FS}, P i PS, i=1,, PS ; P i denotes the nae set for P i, P i ={ k W_P ik =1, k S}, P i PS, i=1,, PS ; W_F k denotes whether F is assigned to k, F FS, k S, =1,, FS, k=1,, S, which is equal to: 1, if the face F is assigned to the nae k W _ Fk 0, otherwise FS_ k denotes the total associated face set for k based on the global assignent of naes and faces, k S, k=1,, S ; LCD(FS_, k) denotes the utual density aong all the faces in the face set FS_ related to the nae under a certain global alignent anner, S, =1,, S. Thus our atheatical odel can be foralized as: ax LCDFS _, k Fi FS _ LDSFi, FS _, k(7) S ax S FS _ FS _ s.t., 1) FPi FS, Pi S. That is, the union set of the 1i PS 1i PS face set included in each pair is the universal face set of the whole database, and the union set of the nae set included in each pair is the universal nae set. 2) FP i FP =Ф, i, i, =1,, PS. That is, there is no intersection set between the face sets of two pairs. 3) FP i = P i, i=1,, PS. That is, to establish ore efficient coding inforation for the optiization algorith, we for a constraint that the nubers of faces and naes in an arbitrary pair ust be sae. ull faces or naes can be used to suppleent the less ones. 4) Pi 1 _ k W Fk 1, F FP i, k P i, i=1,, PS, =1,, FP i. That is, in a pair each real face can only be assigned to FPi at ost a real nae in the sae pair. 5) 1 _ W Fk 1, F FP i, k P i, i=1,, PS, k=1,, P i. That is, in a pair each real nae can only be assigned to at ost a real face in the sae pair. 6), _ F 0 FPi l P W Fl, i=1,, PS, i =1,, FP i, l=1,, S. That is, the face that belongs to a pair can only be assigned to the nae in the sae pair. 6.2 ISSAGA-based Multiodal Optiization a. Encoding of Chroosoe The naes fro all the pairs are positioned in a fixed order, and the faces fro all the pairs are ranked in segents and then grouped into a chroosoe. Each chroosoe corresponds to a solution. A repeatable natural nuber encoding pattern is adopted to design the gene g i of the chroosoe C, C={g i}, i=1,, PS, =1,, FP i, where i is the pair nuber; is the face nuber for the pair P i ; and g i denotes the nuber for the face F in P i. C can be expanded as {g 11,, g 1 FP1,, g i1,, g i FPi,, g PS 1,, g PS FP PS }, where {g i1,, g i FPi } is a segent and each segent keeps the relatively independent relationship with the other segents. b. Initial Population The initial population P(t) with L chroosoes is produced by a stochastic ethod. The stochastic order in each segent is used to generate an initial chroosoe C, and then the other L-1 different chroosoes are achieved fro C. According to the obective function value, the current best chroosoe can be regarded as the initial optial solution. c. Self-Adaptive Selection-Reproduction Operator According to our atheatical odel, the obective function for the l th chroosoe in a population can be constructed as: LCD FS _, k OF Cl (8) S FS _ To liit the quantity of chroosoes that dissatisfy the constraints in the new population, we use a Multiple Roulette Wheel ethod and copute the selection probability as: f ' C (9) l PS Cl, l 1,, M M f ' Ck k1 where M is the nuber of chroosoes in the current population, and f ( ) is obtained through the self-adaptive transforation based on the original fitness function f( ) to avoid preaturity and slow convergence speed. g ax e e f Cl OFCl, f ' Cl a * f Cl g * f ax f in(10) g ax e e where f ax and f in are the axiu and iniu fitness values for the current population; g is the nuber of generations under the current genetic anner; g ax is the axiu nuber of genetic generations; and a is a constant paraeter. d. Self-Adaptive Siulated Annealing Crossover Operator Considering in ost cases there are two or three faces in a news iage and the axiu order nuber for the faces in a segent of a chroosoe is also two or three, thus a One-Point crossover ethod is adopted. Two special strategies of the self-adaptive crossover probability (P C ) coputation and the siulated annealing operation are introduced. PC1 PC 2ax f Ci, f C (11) PC1, ax f Ci, f C PC Ci, C f ax PC1, ax f Ci, f C where f avg is the average fitness value in the current population; P C1 and P C2 are two predefined paraeters. The siulated annealing operation is utilized to udge whether the inferior solution should be accepted to substitute the previous chroosoe when facing with an inferior solution and generate a probability (P A ) for accepting the inferior solution. 1 PA Cl (12) f Cl' f Cl g T 0 1 e where f(c l ) denotes the fitness value for the new chroosoe C l generated after the crossover; T 0 is the initial teperature for the siulated annealing operation; and δ is a preset ratio coefficient for the teperature reduction. e. Self-Adaptive Siulated Annealing Mutation Operator The utation operator cannot be executed aong segents but inside every segent by utilizing an Exchange Mutation ethod. Meanwhile, two special strategies of the self-adaptive utation probability (P M ) coputation and the siulated annealing operation are also introduced. PM 1 PM 2f ax f Cl PM 1, f Cl (13) PM Cl f ax PM 1, f Cl where P M1 and P M2 are two predefined paraeters. Given a chroosoe C k ={S 1,, S i,, S }, = PS, S i is the i th segent. Taking a segent as a unit, the 2-exchange utation is accoplished aong the segents of C k by P M (C k ). g 2771

5 7 Experient and Analysis 7.1 Dataset and Evaluation Metrics Our dataset is established based on Labeled Yahoo! ews Data constructed by Berg et al. [2005] and further developed by Guillauin et al. [2008]. It consists of 20,071 captioned news iages, and all the iage-caption pairs contain 31,147 face iages that belong to 5,873 different persons/naes. To evaluate the effectiveness of our algorith, the ground truth associations are considered to easure the benchark etrics for nae-face aligned pairs. Correct Label Rate (CLR) is defined as the percentage of correct pairs generated in the ground truth links. Our evaluation can not only consider the association links between real naes and faces, but also concern the assignents to null naes or null faces. 7.2 Experient on Cross-Media Alignent Our cross-edia alignent odel is created by integrating ae Salience Ranking (SR), ae-face Cohesion Measure (FCM), Web-based Multiodal Inforation Mining (WMIM) for rare nae and the Iproved Self-Adaptive Siulated Annealing Genetic Algorith of ISSAGA. To investigate the effect of each part on the whole alignent perforance, we introduce three evaluation patterns: 1) Baseline(FCM) based on the basic preprocessing with o SR for captions and FCM for iages; 2) Baseline(FCM)+SR based on the baseline odel with FCM, SR is integrated; 3) Baseline(FCM)+SR+WMIM by fusing the baseline odel with FCM and SR, WMIM is added. These patterns are tested in two alignent optiization anners of general GA and ISSAGA, which ais to explore the advantages of our proposed ISSAGA. To verify the consistency of our alignent odel in different apping cases, the perforance evaluation focuses on two assignent schees. One contains the specific appings of Real Face->ull ae (RF->) and Real ae->ull Face (R->F) except the general apping of Real Face<->Real ae (), and another only involves. The experiental results are shown in Table 1. Assignent Schee Evaluation Pattern GA ISSAGA CLR (%) CLR (%) Baseline(FCM) [o SR] RF-> Baseline(FCM)+SR R->F Baseline(FCM)+SR+WMIM Baseline(FCM) [o SR] Baseline(FCM)+SR Baseline(FCM)+SR+WMIM Table 1. The experiental results on our whole dataset. It can be seen fro Table 1 that for the nae-face linking in our whole dataset, we can still obtain the best CLR value of 65.77% in the sae evaluation pattern of fusing Baseline with FCM, SR, WMIM and ISSAGA and excluding the appings related to ull ae and ull Face. In coparison with the baseline odel, the alignent perforance could be prooted to a great degree by successively adding SR and WMIM under our ISSAGA-based optiization. Copared the results for different assignent schees, we can observe that when only considering the alignent between real naes and faces, the alignent perforance increases and becoes pretty good. Although when evaluating the links with null naes and faces the alignent perforance aybe slightly influenced by such extra uncertainty, the CLR can still reach a relatively high value. Through the coparison between the results based on the traditional GA and ISSAGA, it can be found that our ISSAGA is obviously superior to the traditional GA and ore suitable for the alignent optiization. In addition, the convergence curves of the best solutions found by the traditional GA and our ISSAGA are shown in Figure 1. The traditional GA iproves the solutions very fast within 60 iterations, and reaches a plateau after that. In contrast, our ISSAGA keeps constantly iproving the solutions with ore iterations, gives uch better solutions, and has a stronger ability of escaping fro the local optia. It can be concluded that our ISSAGA provides better solutions within a satisfactory aount of tie and is very useful to contribute better solutions for such a special optiization proble. Figure 1. The convergence curves of the traditional GA and our ISSAGA. 7.3 Experient on Cross-Media ews Retrieval To further explore the applicability of our alignent odel in cross-edia news retrieval, we particularly select 16 naes with the larger nuber of faces in the whole dataset as the query topics and carry out the retrieval runs based on the original dataset and the dataset with the alignent processing respectively. The Precision (P), Recall (R) and F-easure (F) values for each nae query are shown in Table 2. It can be seen that the best run is based on the dataset with alignent processing, and its results exceed those by another run based on the original dataset without any alignent processing draatically. For all the nae queries, their F values under the run with alignent processing are obviously higher than those under another run, and the highest increase can reach to 20%. By adopting our alignent odel for the original dataset, the whole retrieval perforance for the selected nae queries has gained the significant iproveent. Captioned ews Iage Dataset for Testing ae Query The Original Dataset The Dataset with Alignent Processing Baseline+SR SR+FCM+WMIM+ISSAGA P (%) R (%) F (%) P (%) R (%) F (%) George W. Bush Collin Powell Tony Blair Donald Rusfeld Gerhard Schroeder Ariel Sharon Hugo Chavez Junichiro Koizui Jean Chretien John Ashcroft Jacques Chirac Hans Blix Vladiir Putin Silvio Berlusconi Arnold Schwarzenegger John egroponte Table 2. The experiental results for cross-edia retrieval. 2772

6 7.4 Coparison with Existing Approaches To give full exhibition to the superiority of our alignent odel, we have also perfored a coparison between our approach and the other classical ones in recent years. Three approaches developed by Berg et al. [2005], Guillauin et al. [2008] and Pha et al. [2010] are analogous with ours, and then we accoplished the on the sae dataset. The experiental results are presented in Table 3, which reflect the difference of power between these four approaches. Approach Assignent Schee Evaluation Pattern CLR (%) Berg [o SR] RF-> Berg+SR R->F Berg+SR+WMIM Berg et al. s [2005] (Berg) Guillauin et al. s [2008] (Guill) Pha et al. s [2010] (Pha) Our Approach with ISSAGA RF-> R->F RF-> R->F RF-> R->F Berg [o SR] Berg+SR Berg+SR+WMIM Guill [o SR] Guill+SR Guill+SR+WMIM Guill [o SR] Guill+SR Guill+SR+WMIM Pha [o SR] Pha+SR Pha+SR+WMIM Pha [o SR] Pha+SR Pha+SR+WMIM Baseline(FCM) [o SR] Baseline(FCM)+SR Baseline(FCM)+SR+WMIM Baseline(FCM) [o SR] Baseline(FCM)+SR Baseline(FCM)+SR+WMIM Table 3. The coparison between our and the other existing ethods. It can be found fro Table 3 that for the nae-face linking on our whole dataset by Berg/Guiilauin/Pha et al. s approaches, we can obtain the best CLR values of 52.71%, 54.41% and 55.61% in the evaluation patterns of Berg/Guill/ Pha [o SR] and excluding the appings related to ull ae and ull Face respectively. The ain reason is that when only considering the alignents between real faces and real naes, all these three approaches can reduce the cobination abiguities between naes and faces in the sae iage-caption pair to a certain degree and then the relatively better CLR values can be obtained. Copared the results of Berg/Guill/Pha [o SR] and our baseline odel with FCM, we can find the best CLR value of 56.78% appears in the results of our odel, which is obviously higher than those CLR values of 52.71%, 54.41% and 55.61% for Berg/Guill/ Pha [o SR] respectively. This iplies that both of our nae-face cohesion easure echanis FCM and cobination optiization algorith ISSAGA are feasible for facilitating ore effective nae-face alignent evaluation and optiization. Furtherore, through fusing our SR and WMIM with Berg/Guillauin/Pha et al. s approaches in turn, the better perforance can be acquired, in which the best CLR values are considerably iproved to 59.93%, 62.43% and 62.73% respectively. This confirs the proinent roles of our SR and WMIM in autoatic nae-face alignent once again. Copared with the iproved Berg/ Guillauin/Pha et al. s approaches that integrate with SR and WMIM, our alignent odel with FCM, SR, WMIM and ISSAGA can still present the better perforance and the best CLR value of 65.77% differs greatly fro those of Berg/Guillauin/Pha et al. s. This indicates that our approach is really superior to Berg/Guillauin/Pha et al. s, and also further confirs that our alignent odel with FCM, SR, WMIM and ISSAGA is exactly a better way for deterining nae-face associations and can support such a special cobinatorial optiization proble ore effectively. 7.5 Analysis and Discussion Through the analysis for the aligned nae-face linkages with failure, it can be found: 1) The alignent acquisition is associated with the preprocessing for iage-caption pairs, especially for caption text. It s easier for iage and caption preprocessing to produce soe errors or issing detections for faces and naes. Such wrong or undetected inforation will seriously affect the whole alignent perforance. 2) The other naes that co-occur with a certain nae in the sae caption aybe helpful for facilitating this nae s related face finding and naing. The sae case is for faces in iages. Such nae or face co-occurrence can be utilized to further iprove the alignent easureent. 3) For ultiodal inforation fro Web ining, instead of directly using the feedback fro iage search engines, it s helpful to exploit a relevance evaluation anner to select ore appropriate copleentary inforation for rare nae. (4) It is a novel way to solve nae-face alignent fro the view of cobinatorial optiization. Although our ISSAGA have obtained the satisfactory perforance on large-scale dataset, aking ore highly specialized consideration about genetic operations will be ore beneficial to acquiring ore feasible solutions. 5) Soe news captions involve extreely rare nae. With very liited inforation suppleent fro Web, such news lacks enough ultiodal inforation for alignent processing. This ay be a ore stubborn proble. 8 Conclusions and Future Work A new algorith is introduced in this paper to support ore precise autoatic nae-face alignent for cross-edia news retrieval. The ulti-level analysis of iage-caption pairs is established for characterizing the eaningful naes with higher salience and the cohesion between naes and faces. To reedy the issue of rare nae, the extra ultiodal inforation is acquired through Web ining. ISSAGA is introduced to solve the alignent optiization better. Thus a novel odel is developed by integrating all these aspects to effectively exploit nae-face correlations. Our future work will focus on aking our syste available online, so that ore Internet users can benefit fro our research. Acknowledgents This work is supported by ational atural Science Fund of China (o ; o ), ational Science & Technology Pillar Progra of China (o. 2012BAH59F04), 973 Progra of China (o. 2010CB327906), Shanghai Outstanding Acadeic Leader Progra (o. 12XD ), and Shanghai Leading Acadeic Discipline Proect (o. B114). Cheng Jin is the corresponding author. 2773

7 References [Andresen et al., 2008] M. Andresen, H. Bräsel, J. Tusch, M. Mörig, F. Werner, and P. Willenius. Siulated annealing and genetic algoriths for iniizing ean flow tie in an open shop. Matheatical and Coputer Modelling, 48(7-8): , [Berg et al., 2005] T.L. Berg, A.C. Berg, J. Edwards, and D.A. Forsyth. Who s in the picture. Advances in eural Inforation Processing Systes 17, , [Berg et al., 2007] T.L. Berg, A.C. Berg, J. Edwards, and M. Maire. aes and faces. Technical Report, U.C. at Berkeley, [Berg et al., 2004] T.L. Berg, A.C. Berg, J. Edwards, M. Maire, R. White, Y.W. The, E. Learned-Miller, and D. Forsyth. aes and faces in the news. In Proceedings of CVPR 2004, 2: , [Bozorgtabar et al., 2011] B. Bozorgtabar and G.A. Rezai Rad. A genetic prograing - PCA hybrid face recognition algorith. Journal of Signal and Inforation Processing, 2: , [Datta et al., 2008] R. Datta, D. Joshi, J. Li, and J.Z. Wang. Iage retrieval: Ideas, influences, and trends of the new age. ACM Coputing Surveys (CSUR), 40(2), Article 5, [Deschacht et al., 2007] K. Deschacht and M.F. Moens, Text analysis for autoatic iage annotation. In Proceedings of ACL2007, , [Everingha et al., 2006] M. Everingha, J. Sivic, and A. Zisseran. Hello! My nae Is Buffy - Autoatic naing of characters in TV video. In Proceedings of BMVC 2006, , [Everingha et al., 2009] M. Everingha, J. Sivic, and A. Zisseran. Taking the bite out of autoatic naing of characters in TV video. Iage and Vision Coputing, 27(5): , [Fan et al., 2012] J.P. Fan, X.F. He,. Zhou, J.Y. Peng, and R. Jain. Quantitative characterization of seantic gaps for learning coplexity estiation and inference odel selection. IEEE Transactions on Multiedia, 14(5): , [Guillauin et al., 2008] M. Guillauin, T. Mensink, J. Verbeek, and C. Schid. Autoatic face naing with caption-based supervision. In Proceedings of CVPR 2008, 1-8, [Guillauin et al., 2012] M. Guillauin, T. Mensink, J. Verbeek, and C. Schid. Face recognition fro caption-based supervision. International Journal of Coputer Vision, 96(1):64-82, [Le et al., 2008] D.D. Le and S. Satoh. Unsupervised face annotation by ining the Web. In Proceedings of ICDM 2008, , [Le et al., 2007] D.D. Le, S. Satoh, M.E. Houle, and D.P.T. guyen. Finding iportant people in large news video databases using ultiodal and clustering analysis. In Proceedings of ICDEW 2007, , [Liu et al., 2008] C. Liu, S. Jiang, and Q. Huang. aing faces in broadcast news video by iage Google. In Proceedings of MM 2008, , [Mensink et al., 2008] T. Mensink and J. Verbeek. Iproving People Search using Query expansions: How friends help to find people. In Proceedings of ECCV 2008, 86-99, [Ozkan et al., 2006] D. Ozkan and P. Duygulu. A graph based approach for naing faces in news photo. In Proceedings of CVPR 2006, , [Pha et al., 2010] P.T. Pha, M.F. Moens, and T. Tuytelaars. Cross-edia alignent of naes and faces. IEEE Transactions on Multiedia, 12(1):13-27, [Pha et al., 2011] P.T. Pha, T. Tuytelaars, and M.F. Moens. aing people in news videos with label propagation. IEEE Multiedia, 18(3): 44-55, [Satoh et al., 1997] S. Satoh and T. Kanade. ae-it: Association of face and nae in video. In Proceedings of CVPR 1997, , [Satoh et al., 1999] S. Satoh, Y. akaura, and T. Kanade. ae-it: aing and Detecting Faces in ews Videos. IEEE Multiedia, 6(1):22-35, [Wang et al., 2005] Z.G. Wang, M. Rahan., and Y.S. Wong. Optiization of ulti-pass illing using parallel genetic algorith and parallel genetic siulated annealing. International Journal of Machine Tools and Manufacture, 45(15): , [Yang et al., 2004] J. Yang, M.Y. Chen, and A.G. Hauptann. Finding person x: Correlating naes with visual appearances. In Proceedings of CIVR 2004, , [Yang et al., 2009] Y. Yang, D. Xu, F.P. ie, J.B. Luo, and Y.T. Zhuang. Ranking with local regression and global alignent for cross edia retrieval. In Proceedings of MM 2009, ,

Applying Multiple Neural Networks on Large Scale Data

Applying Multiple Neural Networks on Large Scale Data 0 International Conference on Inforation and Electronics Engineering IPCSIT vol6 (0) (0) IACSIT Press, Singapore Applying Multiple Neural Networks on Large Scale Data Kritsanatt Boonkiatpong and Sukree

More information

PERFORMANCE METRICS FOR THE IT SERVICES PORTFOLIO

PERFORMANCE METRICS FOR THE IT SERVICES PORTFOLIO Bulletin of the Transilvania University of Braşov Series I: Engineering Sciences Vol. 4 (53) No. - 0 PERFORMANCE METRICS FOR THE IT SERVICES PORTFOLIO V. CAZACU I. SZÉKELY F. SANDU 3 T. BĂLAN Abstract:

More information

An Innovate Dynamic Load Balancing Algorithm Based on Task

An Innovate Dynamic Load Balancing Algorithm Based on Task An Innovate Dynaic Load Balancing Algorith Based on Task Classification Hong-bin Wang,,a, Zhi-yi Fang, b, Guan-nan Qu,*,c, Xiao-dan Ren,d College of Coputer Science and Technology, Jilin University, Changchun

More information

International Journal of Management & Information Systems First Quarter 2012 Volume 16, Number 1

International Journal of Management & Information Systems First Quarter 2012 Volume 16, Number 1 International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 Proposal And Effectiveness Of A Highly Copelling Direct Mail Method - Establishent And Deployent Of PMOS-DM Hisatoshi

More information

An Approach to Combating Free-riding in Peer-to-Peer Networks

An Approach to Combating Free-riding in Peer-to-Peer Networks An Approach to Cobating Free-riding in Peer-to-Peer Networks Victor Ponce, Jie Wu, and Xiuqi Li Departent of Coputer Science and Engineering Florida Atlantic University Boca Raton, FL 33431 April 7, 2008

More information

Fuzzy Sets in HR Management

Fuzzy Sets in HR Management Acta Polytechnica Hungarica Vol. 8, No. 3, 2011 Fuzzy Sets in HR Manageent Blanka Zeková AXIOM SW, s.r.o., 760 01 Zlín, Czech Republic blanka.zekova@sezna.cz Jana Talašová Faculty of Science, Palacký Univerzity,

More information

Analyzing Spatiotemporal Characteristics of Education Network Traffic with Flexible Multiscale Entropy

Analyzing Spatiotemporal Characteristics of Education Network Traffic with Flexible Multiscale Entropy Vol. 9, No. 5 (2016), pp.303-312 http://dx.doi.org/10.14257/ijgdc.2016.9.5.26 Analyzing Spatioteporal Characteristics of Education Network Traffic with Flexible Multiscale Entropy Chen Yang, Renjie Zhou

More information

The Research of Measuring Approach and Energy Efficiency for Hadoop Periodic Jobs

The Research of Measuring Approach and Energy Efficiency for Hadoop Periodic Jobs Send Orders for Reprints to reprints@benthascience.ae 206 The Open Fuels & Energy Science Journal, 2015, 8, 206-210 Open Access The Research of Measuring Approach and Energy Efficiency for Hadoop Periodic

More information

Media Adaptation Framework in Biofeedback System for Stroke Patient Rehabilitation

Media Adaptation Framework in Biofeedback System for Stroke Patient Rehabilitation Media Adaptation Fraework in Biofeedback Syste for Stroke Patient Rehabilitation Yinpeng Chen, Weiwei Xu, Hari Sundara, Thanassis Rikakis, Sheng-Min Liu Arts, Media and Engineering Progra Arizona State

More information

Managing Complex Network Operation with Predictive Analytics

Managing Complex Network Operation with Predictive Analytics Managing Coplex Network Operation with Predictive Analytics Zhenyu Huang, Pak Chung Wong, Patrick Mackey, Yousu Chen, Jian Ma, Kevin Schneider, and Frank L. Greitzer Pacific Northwest National Laboratory

More information

The AGA Evaluating Model of Customer Loyalty Based on E-commerce Environment

The AGA Evaluating Model of Customer Loyalty Based on E-commerce Environment 6 JOURNAL OF SOFTWARE, VOL. 4, NO. 3, MAY 009 The AGA Evaluating Model of Custoer Loyalty Based on E-coerce Environent Shaoei Yang Econoics and Manageent Departent, North China Electric Power University,

More information

CLOSED-LOOP SUPPLY CHAIN NETWORK OPTIMIZATION FOR HONG KONG CARTRIDGE RECYCLING INDUSTRY

CLOSED-LOOP SUPPLY CHAIN NETWORK OPTIMIZATION FOR HONG KONG CARTRIDGE RECYCLING INDUSTRY CLOSED-LOOP SUPPLY CHAIN NETWORK OPTIMIZATION FOR HONG KONG CARTRIDGE RECYCLING INDUSTRY Y. T. Chen Departent of Industrial and Systes Engineering Hong Kong Polytechnic University, Hong Kong yongtong.chen@connect.polyu.hk

More information

Software Quality Characteristics Tested For Mobile Application Development

Software Quality Characteristics Tested For Mobile Application Development Thesis no: MGSE-2015-02 Software Quality Characteristics Tested For Mobile Application Developent Literature Review and Epirical Survey WALEED ANWAR Faculty of Coputing Blekinge Institute of Technology

More information

INTEGRATED ENVIRONMENT FOR STORING AND HANDLING INFORMATION IN TASKS OF INDUCTIVE MODELLING FOR BUSINESS INTELLIGENCE SYSTEMS

INTEGRATED ENVIRONMENT FOR STORING AND HANDLING INFORMATION IN TASKS OF INDUCTIVE MODELLING FOR BUSINESS INTELLIGENCE SYSTEMS Artificial Intelligence Methods and Techniques for Business and Engineering Applications 210 INTEGRATED ENVIRONMENT FOR STORING AND HANDLING INFORMATION IN TASKS OF INDUCTIVE MODELLING FOR BUSINESS INTELLIGENCE

More information

arxiv:0805.1434v1 [math.pr] 9 May 2008

arxiv:0805.1434v1 [math.pr] 9 May 2008 Degree-distribution stability of scale-free networs Zhenting Hou, Xiangxing Kong, Dinghua Shi,2, and Guanrong Chen 3 School of Matheatics, Central South University, Changsha 40083, China 2 Departent of

More information

Machine Learning Applications in Grid Computing

Machine Learning Applications in Grid Computing Machine Learning Applications in Grid Coputing George Cybenko, Guofei Jiang and Daniel Bilar Thayer School of Engineering Dartouth College Hanover, NH 03755, USA gvc@dartouth.edu, guofei.jiang@dartouth.edu

More information

An Integrated Approach for Monitoring Service Level Parameters of Software-Defined Networking

An Integrated Approach for Monitoring Service Level Parameters of Software-Defined Networking International Journal of Future Generation Counication and Networking Vol. 8, No. 6 (15), pp. 197-4 http://d.doi.org/1.1457/ijfgcn.15.8.6.19 An Integrated Approach for Monitoring Service Level Paraeters

More information

Image restoration for a rectangular poor-pixels detector

Image restoration for a rectangular poor-pixels detector Iage restoration for a rectangular poor-pixels detector Pengcheng Wen 1, Xiangjun Wang 1, Hong Wei 2 1 State Key Laboratory of Precision Measuring Technology and Instruents, Tianjin University, China 2

More information

Real Time Target Tracking with Binary Sensor Networks and Parallel Computing

Real Time Target Tracking with Binary Sensor Networks and Parallel Computing Real Tie Target Tracking with Binary Sensor Networks and Parallel Coputing Hong Lin, John Rushing, Sara J. Graves, Steve Tanner, and Evans Criswell Abstract A parallel real tie data fusion and target tracking

More information

Use of extrapolation to forecast the working capital in the mechanical engineering companies

Use of extrapolation to forecast the working capital in the mechanical engineering companies ECONTECHMOD. AN INTERNATIONAL QUARTERLY JOURNAL 2014. Vol. 1. No. 1. 23 28 Use of extrapolation to forecast the working capital in the echanical engineering copanies A. Cherep, Y. Shvets Departent of finance

More information

Online Bagging and Boosting

Online Bagging and Boosting Abstract Bagging and boosting are two of the ost well-known enseble learning ethods due to their theoretical perforance guarantees and strong experiental results. However, these algoriths have been used

More information

Extended-Horizon Analysis of Pressure Sensitivities for Leak Detection in Water Distribution Networks: Application to the Barcelona Network

Extended-Horizon Analysis of Pressure Sensitivities for Leak Detection in Water Distribution Networks: Application to the Barcelona Network 2013 European Control Conference (ECC) July 17-19, 2013, Zürich, Switzerland. Extended-Horizon Analysis of Pressure Sensitivities for Leak Detection in Water Distribution Networks: Application to the Barcelona

More information

Searching strategy for multi-target discovery in wireless networks

Searching strategy for multi-target discovery in wireless networks Searching strategy for ulti-target discovery in wireless networks Zhao Cheng, Wendi B. Heinzelan Departent of Electrical and Coputer Engineering University of Rochester Rochester, NY 467 (585) 75-{878,

More information

ASIC Design Project Management Supported by Multi Agent Simulation

ASIC Design Project Management Supported by Multi Agent Simulation ASIC Design Project Manageent Supported by Multi Agent Siulation Jana Blaschke, Christian Sebeke, Wolfgang Rosenstiel Abstract The coplexity of Application Specific Integrated Circuits (ASICs) is continuously

More information

Evaluating Inventory Management Performance: a Preliminary Desk-Simulation Study Based on IOC Model

Evaluating Inventory Management Performance: a Preliminary Desk-Simulation Study Based on IOC Model Evaluating Inventory Manageent Perforance: a Preliinary Desk-Siulation Study Based on IOC Model Flora Bernardel, Roberto Panizzolo, and Davide Martinazzo Abstract The focus of this study is on preliinary

More information

Efficient Key Management for Secure Group Communications with Bursty Behavior

Efficient Key Management for Secure Group Communications with Bursty Behavior Efficient Key Manageent for Secure Group Counications with Bursty Behavior Xukai Zou, Byrav Raaurthy Departent of Coputer Science and Engineering University of Nebraska-Lincoln Lincoln, NE68588, USA Eail:

More information

Exercise 4 INVESTIGATION OF THE ONE-DEGREE-OF-FREEDOM SYSTEM

Exercise 4 INVESTIGATION OF THE ONE-DEGREE-OF-FREEDOM SYSTEM Eercise 4 IVESTIGATIO OF THE OE-DEGREE-OF-FREEDOM SYSTEM 1. Ai of the eercise Identification of paraeters of the euation describing a one-degree-of- freedo (1 DOF) atheatical odel of the real vibrating

More information

High Performance Chinese/English Mixed OCR with Character Level Language Identification

High Performance Chinese/English Mixed OCR with Character Level Language Identification 2009 0th International Conference on Docuent Analysis and Recognition High Perforance Chinese/English Mixed OCR with Character Level Language Identification Kai Wang Institute of Machine Intelligence,

More information

An improved TF-IDF approach for text classification *

An improved TF-IDF approach for text classification * Zhang et al. / J Zheiang Univ SCI 2005 6A(1:49-55 49 Journal of Zheiang University SCIECE ISS 1009-3095 http://www.zu.edu.cn/zus E-ail: zus@zu.edu.cn An iproved TF-IDF approach for text classification

More information

Preference-based Search and Multi-criteria Optimization

Preference-based Search and Multi-criteria Optimization Fro: AAAI-02 Proceedings. Copyright 2002, AAAI (www.aaai.org). All rights reserved. Preference-based Search and Multi-criteria Optiization Ulrich Junker ILOG 1681, route des Dolines F-06560 Valbonne ujunker@ilog.fr

More information

Standards and Protocols for the Collection and Dissemination of Graduating Student Initial Career Outcomes Information For Undergraduates

Standards and Protocols for the Collection and Dissemination of Graduating Student Initial Career Outcomes Information For Undergraduates National Association of Colleges and Eployers Standards and Protocols for the Collection and Disseination of Graduating Student Initial Career Outcoes Inforation For Undergraduates Developed by the NACE

More information

A framework for performance monitoring, load balancing, adaptive timeouts and quality of service in digital libraries

A framework for performance monitoring, load balancing, adaptive timeouts and quality of service in digital libraries Int J Digit Libr (2000) 3: 9 35 INTERNATIONAL JOURNAL ON Digital Libraries Springer-Verlag 2000 A fraework for perforance onitoring, load balancing, adaptive tieouts and quality of service in digital libraries

More information

Optimal Resource-Constraint Project Scheduling with Overlapping Modes

Optimal Resource-Constraint Project Scheduling with Overlapping Modes Optial Resource-Constraint Proect Scheduling with Overlapping Modes François Berthaut Lucas Grèze Robert Pellerin Nathalie Perrier Adnène Hai February 20 CIRRELT-20-09 Bureaux de Montréal : Bureaux de

More information

Design of Model Reference Self Tuning Mechanism for PID like Fuzzy Controller

Design of Model Reference Self Tuning Mechanism for PID like Fuzzy Controller Research Article International Journal of Current Engineering and Technology EISSN 77 46, PISSN 347 56 4 INPRESSCO, All Rights Reserved Available at http://inpressco.co/category/ijcet Design of Model Reference

More information

Fuzzy Evaluation on Network Security Based on the New Algorithm of Membership Degree Transformation M(1,2,3)

Fuzzy Evaluation on Network Security Based on the New Algorithm of Membership Degree Transformation M(1,2,3) 324 JOURNAL OF NETWORKS, VOL. 4, NO. 5, JULY 29 Fuzzy Evaluation on Networ Security Based on the New Algorith of Mebership Degree Transforation M(,2,3) Hua Jiang School of Econoics and Manageent, Hebei

More information

Exploiting Hardware Heterogeneity within the Same Instance Type of Amazon EC2

Exploiting Hardware Heterogeneity within the Same Instance Type of Amazon EC2 Exploiting Hardware Heterogeneity within the Sae Instance Type of Aazon EC2 Zhonghong Ou, Hao Zhuang, Jukka K. Nurinen, Antti Ylä-Jääski, Pan Hui Aalto University, Finland; Deutsch Teleko Laboratories,

More information

Evaluating the Effectiveness of Task Overlapping as a Risk Response Strategy in Engineering Projects

Evaluating the Effectiveness of Task Overlapping as a Risk Response Strategy in Engineering Projects Evaluating the Effectiveness of Task Overlapping as a Risk Response Strategy in Engineering Projects Lucas Grèze Robert Pellerin Nathalie Perrier Patrice Leclaire February 2011 CIRRELT-2011-11 Bureaux

More information

This paper studies a rental firm that offers reusable products to price- and quality-of-service sensitive

This paper studies a rental firm that offers reusable products to price- and quality-of-service sensitive MANUFACTURING & SERVICE OPERATIONS MANAGEMENT Vol., No. 3, Suer 28, pp. 429 447 issn 523-464 eissn 526-5498 8 3 429 infors doi.287/so.7.8 28 INFORMS INFORMS holds copyright to this article and distributed

More information

Research Article Performance Evaluation of Human Resource Outsourcing in Food Processing Enterprises

Research Article Performance Evaluation of Human Resource Outsourcing in Food Processing Enterprises Advance Journal of Food Science and Technology 9(2): 964-969, 205 ISSN: 2042-4868; e-issn: 2042-4876 205 Maxwell Scientific Publication Corp. Subitted: August 0, 205 Accepted: Septeber 3, 205 Published:

More information

AN ALGORITHM FOR REDUCING THE DIMENSION AND SIZE OF A SAMPLE FOR DATA EXPLORATION PROCEDURES

AN ALGORITHM FOR REDUCING THE DIMENSION AND SIZE OF A SAMPLE FOR DATA EXPLORATION PROCEDURES Int. J. Appl. Math. Coput. Sci., 2014, Vol. 24, No. 1, 133 149 DOI: 10.2478/acs-2014-0011 AN ALGORITHM FOR REDUCING THE DIMENSION AND SIZE OF A SAMPLE FOR DATA EXPLORATION PROCEDURES PIOTR KULCZYCKI,,

More information

CRM FACTORS ASSESSMENT USING ANALYTIC HIERARCHY PROCESS

CRM FACTORS ASSESSMENT USING ANALYTIC HIERARCHY PROCESS 641 CRM FACTORS ASSESSMENT USING ANALYTIC HIERARCHY PROCESS Marketa Zajarosova 1* *Ph.D. VSB - Technical University of Ostrava, THE CZECH REPUBLIC arketa.zajarosova@vsb.cz Abstract Custoer relationship

More information

Resource Allocation in Wireless Networks with Multiple Relays

Resource Allocation in Wireless Networks with Multiple Relays Resource Allocation in Wireless Networks with Multiple Relays Kağan Bakanoğlu, Stefano Toasin, Elza Erkip Departent of Electrical and Coputer Engineering, Polytechnic Institute of NYU, Brooklyn, NY, 0

More information

Method of supply chain optimization in E-commerce

Method of supply chain optimization in E-commerce MPRA Munich Personal RePEc Archive Method of supply chain optiization in E-coerce Petr Suchánek and Robert Bucki Silesian University - School of Business Adinistration, The College of Inforatics and Manageent

More information

Modeling Parallel Applications Performance on Heterogeneous Systems

Modeling Parallel Applications Performance on Heterogeneous Systems Modeling Parallel Applications Perforance on Heterogeneous Systes Jaeela Al-Jaroodi, Nader Mohaed, Hong Jiang and David Swanson Departent of Coputer Science and Engineering University of Nebraska Lincoln

More information

AUC Optimization vs. Error Rate Minimization

AUC Optimization vs. Error Rate Minimization AUC Optiization vs. Error Rate Miniization Corinna Cortes and Mehryar Mohri AT&T Labs Research 180 Park Avenue, Florha Park, NJ 0793, USA {corinna, ohri}@research.att.co Abstract The area under an ROC

More information

Quality evaluation of the model-based forecasts of implied volatility index

Quality evaluation of the model-based forecasts of implied volatility index Quality evaluation of the odel-based forecasts of iplied volatility index Katarzyna Łęczycka 1 Abstract Influence of volatility on financial arket forecasts is very high. It appears as a specific factor

More information

( C) CLASS 10. TEMPERATURE AND ATOMS

( C) CLASS 10. TEMPERATURE AND ATOMS CLASS 10. EMPERAURE AND AOMS 10.1. INRODUCION Boyle s understanding of the pressure-volue relationship for gases occurred in the late 1600 s. he relationships between volue and teperature, and between

More information

Cooperative Caching for Adaptive Bit Rate Streaming in Content Delivery Networks

Cooperative Caching for Adaptive Bit Rate Streaming in Content Delivery Networks Cooperative Caching for Adaptive Bit Rate Streaing in Content Delivery Networs Phuong Luu Vo Departent of Coputer Science and Engineering, International University - VNUHCM, Vietna vtlphuong@hciu.edu.vn

More information

The Application of Bandwidth Optimization Technique in SLA Negotiation Process

The Application of Bandwidth Optimization Technique in SLA Negotiation Process The Application of Bandwidth Optiization Technique in SLA egotiation Process Srecko Krile University of Dubrovnik Departent of Electrical Engineering and Coputing Cira Carica 4, 20000 Dubrovnik, Croatia

More information

An Optimal Task Allocation Model for System Cost Analysis in Heterogeneous Distributed Computing Systems: A Heuristic Approach

An Optimal Task Allocation Model for System Cost Analysis in Heterogeneous Distributed Computing Systems: A Heuristic Approach An Optial Tas Allocation Model for Syste Cost Analysis in Heterogeneous Distributed Coputing Systes: A Heuristic Approach P. K. Yadav Central Building Research Institute, Rooree- 247667, Uttarahand (INDIA)

More information

Reliability Constrained Packet-sizing for Linear Multi-hop Wireless Networks

Reliability Constrained Packet-sizing for Linear Multi-hop Wireless Networks Reliability Constrained acket-sizing for inear Multi-hop Wireless Networks Ning Wen, and Randall A. Berry Departent of Electrical Engineering and Coputer Science Northwestern University, Evanston, Illinois

More information

Physics 211: Lab Oscillations. Simple Harmonic Motion.

Physics 211: Lab Oscillations. Simple Harmonic Motion. Physics 11: Lab Oscillations. Siple Haronic Motion. Reading Assignent: Chapter 15 Introduction: As we learned in class, physical systes will undergo an oscillatory otion, when displaced fro a stable equilibriu.

More information

Load Control for Overloaded MPLS/DiffServ Networks during SLA Negotiation

Load Control for Overloaded MPLS/DiffServ Networks during SLA Negotiation Int J Counications, Network and Syste Sciences, 29, 5, 422-432 doi:14236/ijcns292547 Published Online August 29 (http://wwwscirporg/journal/ijcns/) Load Control for Overloaded MPLS/DiffServ Networks during

More information

Adaptive Modulation and Coding for Unmanned Aerial Vehicle (UAV) Radio Channel

Adaptive Modulation and Coding for Unmanned Aerial Vehicle (UAV) Radio Channel Recent Advances in Counications Adaptive odulation and Coding for Unanned Aerial Vehicle (UAV) Radio Channel Airhossein Fereidountabar,Gian Carlo Cardarilli, Rocco Fazzolari,Luca Di Nunzio Abstract In

More information

How To Get A Loan From A Bank For Free

How To Get A Loan From A Bank For Free Finance 111 Finance We have to work with oney every day. While balancing your checkbook or calculating your onthly expenditures on espresso requires only arithetic, when we start saving, planning for retireent,

More information

Factored Models for Probabilistic Modal Logic

Factored Models for Probabilistic Modal Logic Proceedings of the Twenty-Third AAAI Conference on Artificial Intelligence (2008 Factored Models for Probabilistic Modal Logic Afsaneh Shirazi and Eyal Air Coputer Science Departent, University of Illinois

More information

Markovian inventory policy with application to the paper industry

Markovian inventory policy with application to the paper industry Coputers and Cheical Engineering 26 (2002) 1399 1413 www.elsevier.co/locate/copcheeng Markovian inventory policy with application to the paper industry K. Karen Yin a, *, Hu Liu a,1, Neil E. Johnson b,2

More information

Impact of Processing Costs on Service Chain Placement in Network Functions Virtualization

Impact of Processing Costs on Service Chain Placement in Network Functions Virtualization Ipact of Processing Costs on Service Chain Placeent in Network Functions Virtualization Marco Savi, Massio Tornatore, Giacoo Verticale Dipartiento di Elettronica, Inforazione e Bioingegneria, Politecnico

More information

RECURSIVE DYNAMIC PROGRAMMING: HEURISTIC RULES, BOUNDING AND STATE SPACE REDUCTION. Henrik Kure

RECURSIVE DYNAMIC PROGRAMMING: HEURISTIC RULES, BOUNDING AND STATE SPACE REDUCTION. Henrik Kure RECURSIVE DYNAMIC PROGRAMMING: HEURISTIC RULES, BOUNDING AND STATE SPACE REDUCTION Henrik Kure Dina, Danish Inforatics Network In the Agricultural Sciences Royal Veterinary and Agricultural University

More information

Dynamic Placement for Clustered Web Applications

Dynamic Placement for Clustered Web Applications Dynaic laceent for Clustered Web Applications A. Karve, T. Kibrel, G. acifici, M. Spreitzer, M. Steinder, M. Sviridenko, and A. Tantawi IBM T.J. Watson Research Center {karve,kibrel,giovanni,spreitz,steinder,sviri,tantawi}@us.ib.co

More information

Energy Proportionality for Disk Storage Using Replication

Energy Proportionality for Disk Storage Using Replication Energy Proportionality for Disk Storage Using Replication Jinoh Ki and Doron Rote Lawrence Berkeley National Laboratory University of California, Berkeley, CA 94720 {jinohki,d rote}@lbl.gov Abstract Energy

More information

Data Set Generation for Rectangular Placement Problems

Data Set Generation for Rectangular Placement Problems Data Set Generation for Rectangular Placeent Probles Christine L. Valenzuela (Muford) Pearl Y. Wang School of Coputer Science & Inforatics Departent of Coputer Science MS 4A5 Cardiff University George

More information

An Improved Decision-making Model of Human Resource Outsourcing Based on Internet Collaboration

An Improved Decision-making Model of Human Resource Outsourcing Based on Internet Collaboration International Journal of Hybrid Inforation Technology, pp. 339-350 http://dx.doi.org/10.14257/hit.2016.9.4.28 An Iproved Decision-aking Model of Huan Resource Outsourcing Based on Internet Collaboration

More information

Generating Certification Authority Authenticated Public Keys in Ad Hoc Networks

Generating Certification Authority Authenticated Public Keys in Ad Hoc Networks SECURITY AND COMMUNICATION NETWORKS Published online in Wiley InterScience (www.interscience.wiley.co). Generating Certification Authority Authenticated Public Keys in Ad Hoc Networks G. Kounga 1, C. J.

More information

PREDICTION OF POSSIBLE CONGESTIONS IN SLA CREATION PROCESS

PREDICTION OF POSSIBLE CONGESTIONS IN SLA CREATION PROCESS PREDICTIO OF POSSIBLE COGESTIOS I SLA CREATIO PROCESS Srećko Krile University of Dubrovnik Departent of Electrical Engineering and Coputing Cira Carica 4, 20000 Dubrovnik, Croatia Tel +385 20 445-739,

More information

Markov Models and Their Use for Calculations of Important Traffic Parameters of Contact Center

Markov Models and Their Use for Calculations of Important Traffic Parameters of Contact Center Markov Models and Their Use for Calculations of Iportant Traffic Paraeters of Contact Center ERIK CHROMY, JAN DIEZKA, MATEJ KAVACKY Institute of Telecounications Slovak University of Technology Bratislava

More information

Information Processing Letters

Information Processing Letters Inforation Processing Letters 111 2011) 178 183 Contents lists available at ScienceDirect Inforation Processing Letters www.elsevier.co/locate/ipl Offline file assignents for online load balancing Paul

More information

Energy Efficient VM Scheduling for Cloud Data Centers: Exact allocation and migration algorithms

Energy Efficient VM Scheduling for Cloud Data Centers: Exact allocation and migration algorithms Energy Efficient VM Scheduling for Cloud Data Centers: Exact allocation and igration algoriths Chaia Ghribi, Makhlouf Hadji and Djaal Zeghlache Institut Mines-Téléco, Téléco SudParis UMR CNRS 5157 9, Rue

More information

6. Time (or Space) Series Analysis

6. Time (or Space) Series Analysis ATM 55 otes: Tie Series Analysis - Section 6a Page 8 6. Tie (or Space) Series Analysis In this chapter we will consider soe coon aspects of tie series analysis including autocorrelation, statistical prediction,

More information

REQUIREMENTS FOR A COMPUTER SCIENCE CURRICULUM EMPHASIZING INFORMATION TECHNOLOGY SUBJECT AREA: CURRICULUM ISSUES

REQUIREMENTS FOR A COMPUTER SCIENCE CURRICULUM EMPHASIZING INFORMATION TECHNOLOGY SUBJECT AREA: CURRICULUM ISSUES REQUIREMENTS FOR A COMPUTER SCIENCE CURRICULUM EMPHASIZING INFORMATION TECHNOLOGY SUBJECT AREA: CURRICULUM ISSUES Charles Reynolds Christopher Fox reynolds @cs.ju.edu fox@cs.ju.edu Departent of Coputer

More information

Investing in corporate bonds?

Investing in corporate bonds? Investing in corporate bonds? This independent guide fro the Australian Securities and Investents Coission (ASIC) can help you look past the return and assess the risks of corporate bonds. If you re thinking

More information

A CHAOS MODEL OF SUBHARMONIC OSCILLATIONS IN CURRENT MODE PWM BOOST CONVERTERS

A CHAOS MODEL OF SUBHARMONIC OSCILLATIONS IN CURRENT MODE PWM BOOST CONVERTERS A CHAOS MODEL OF SUBHARMONIC OSCILLATIONS IN CURRENT MODE PWM BOOST CONVERTERS Isaac Zafrany and Sa BenYaakov Departent of Electrical and Coputer Engineering BenGurion University of the Negev P. O. Box

More information

Considerations on Distributed Load Balancing for Fully Heterogeneous Machines: Two Particular Cases

Considerations on Distributed Load Balancing for Fully Heterogeneous Machines: Two Particular Cases Considerations on Distributed Load Balancing for Fully Heterogeneous Machines: Two Particular Cases Nathanaël Cheriere Departent of Coputer Science ENS Rennes Rennes, France nathanael.cheriere@ens-rennes.fr

More information

Performance Evaluation of Machine Learning Techniques using Software Cost Drivers

Performance Evaluation of Machine Learning Techniques using Software Cost Drivers Perforance Evaluation of Machine Learning Techniques using Software Cost Drivers Manas Gaur Departent of Coputer Engineering, Delhi Technological University Delhi, India ABSTRACT There is a treendous rise

More information

Audio Engineering Society. Convention Paper. Presented at the 119th Convention 2005 October 7 10 New York, New York USA

Audio Engineering Society. Convention Paper. Presented at the 119th Convention 2005 October 7 10 New York, New York USA Audio Engineering Society Convention Paper Presented at the 119th Convention 2005 October 7 10 New York, New York USA This convention paper has been reproduced fro the authors advance anuscript, without

More information

Mathematical Model for Glucose-Insulin Regulatory System of Diabetes Mellitus

Mathematical Model for Glucose-Insulin Regulatory System of Diabetes Mellitus Advances in Applied Matheatical Biosciences. ISSN 8-998 Volue, Nuber (0), pp. 9- International Research Publication House http://www.irphouse.co Matheatical Model for Glucose-Insulin Regulatory Syste of

More information

Investing in corporate bonds?

Investing in corporate bonds? Investing in corporate bonds? This independent guide fro the Australian Securities and Investents Coission (ASIC) can help you look past the return and assess the risks of corporate bonds. If you re thinking

More information

Work Travel and Decision Probling in the Network Marketing World

Work Travel and Decision Probling in the Network Marketing World TRB Paper No. 03-4348 WORK TRAVEL MODE CHOICE MODELING USING DATA MINING: DECISION TREES AND NEURAL NETWORKS Chi Xie Research Assistant Departent of Civil and Environental Engineering University of Massachusetts,

More information

2. FINDING A SOLUTION

2. FINDING A SOLUTION The 7 th Balan Conference on Operational Research BACOR 5 Constanta, May 5, Roania OPTIMAL TIME AND SPACE COMPLEXITY ALGORITHM FOR CONSTRUCTION OF ALL BINARY TREES FROM PRE-ORDER AND POST-ORDER TRAVERSALS

More information

Approximately-Perfect Hashing: Improving Network Throughput through Efficient Off-chip Routing Table Lookup

Approximately-Perfect Hashing: Improving Network Throughput through Efficient Off-chip Routing Table Lookup Approxiately-Perfect ing: Iproving Network Throughput through Efficient Off-chip Routing Table Lookup Zhuo Huang, Jih-Kwon Peir, Shigang Chen Departent of Coputer & Inforation Science & Engineering, University

More information

Protecting Small Keys in Authentication Protocols for Wireless Sensor Networks

Protecting Small Keys in Authentication Protocols for Wireless Sensor Networks Protecting Sall Keys in Authentication Protocols for Wireless Sensor Networks Kalvinder Singh Australia Developent Laboratory, IBM and School of Inforation and Counication Technology, Griffith University

More information

Entity Search Engine: Towards Agile Best-Effort Information Integration over the Web

Entity Search Engine: Towards Agile Best-Effort Information Integration over the Web Entity Search Engine: Towards Agile Best-Effort Inforation Integration over the Web Tao Cheng, Kevin Chen-Chuan Chang University of Illinois at Urbana-Chapaign {tcheng3, kcchang}@cs.uiuc.edu. INTRODUCTION

More information

A Scalable Application Placement Controller for Enterprise Data Centers

A Scalable Application Placement Controller for Enterprise Data Centers W WWW 7 / Track: Perforance and Scalability A Scalable Application Placeent Controller for Enterprise Data Centers Chunqiang Tang, Malgorzata Steinder, Michael Spreitzer, and Giovanni Pacifici IBM T.J.

More information

SOME APPLICATIONS OF FORECASTING Prof. Thomas B. Fomby Department of Economics Southern Methodist University May 2008

SOME APPLICATIONS OF FORECASTING Prof. Thomas B. Fomby Department of Economics Southern Methodist University May 2008 SOME APPLCATONS OF FORECASTNG Prof. Thoas B. Foby Departent of Econoics Southern Methodist University May 8 To deonstrate the usefulness of forecasting ethods this note discusses four applications of forecasting

More information

Introduction to the Microsoft Sync Framework. Michael Clark Development Manager Microsoft

Introduction to the Microsoft Sync Framework. Michael Clark Development Manager Microsoft Introduction to the Michael Clark Developent Manager Microsoft Agenda Why Is Sync both Interesting and Hard Sync Fraework Overview Using the Sync Fraework Future Directions Suary Why Is Sync Iportant Coputing

More information

The Benefit of SMT in the Multi-Core Era: Flexibility towards Degrees of Thread-Level Parallelism

The Benefit of SMT in the Multi-Core Era: Flexibility towards Degrees of Thread-Level Parallelism The enefit of SMT in the Multi-Core Era: Flexibility towards Degrees of Thread-Level Parallelis Stijn Eyeran Lieven Eeckhout Ghent University, elgiu Stijn.Eyeran@elis.UGent.be, Lieven.Eeckhout@elis.UGent.be

More information

Partitioned Elias-Fano Indexes

Partitioned Elias-Fano Indexes Partitioned Elias-ano Indexes Giuseppe Ottaviano ISTI-CNR, Pisa giuseppe.ottaviano@isti.cnr.it Rossano Venturini Dept. of Coputer Science, University of Pisa rossano@di.unipi.it ABSTRACT The Elias-ano

More information

ADJUSTING FOR QUALITY CHANGE

ADJUSTING FOR QUALITY CHANGE ADJUSTING FOR QUALITY CHANGE 7 Introduction 7.1 The easureent of changes in the level of consuer prices is coplicated by the appearance and disappearance of new and old goods and services, as well as changes

More information

ON SELF-ROUTING IN CLOS CONNECTION NETWORKS. BARRY G. DOUGLASS Electrical Engineering Department Texas A&M University College Station, TX 77843-3128

ON SELF-ROUTING IN CLOS CONNECTION NETWORKS. BARRY G. DOUGLASS Electrical Engineering Department Texas A&M University College Station, TX 77843-3128 ON SELF-ROUTING IN CLOS CONNECTION NETWORKS BARRY G. DOUGLASS Electrical Engineering Departent Texas A&M University College Station, TX 778-8 A. YAVUZ ORUÇ Electrical Engineering Departent and Institute

More information

Pricing Asian Options using Monte Carlo Methods

Pricing Asian Options using Monte Carlo Methods U.U.D.M. Project Report 9:7 Pricing Asian Options using Monte Carlo Methods Hongbin Zhang Exaensarbete i ateatik, 3 hp Handledare och exainator: Johan Tysk Juni 9 Departent of Matheatics Uppsala University

More information

Modeling Cooperative Gene Regulation Using Fast Orthogonal Search

Modeling Cooperative Gene Regulation Using Fast Orthogonal Search 8 The Open Bioinforatics Journal, 28, 2, 8-89 Open Access odeling Cooperative Gene Regulation Using Fast Orthogonal Search Ian inz* and ichael J. Korenberg* Departent of Electrical and Coputer Engineering,

More information

A Study on the Chain Restaurants Dynamic Negotiation Games of the Optimization of Joint Procurement of Food Materials

A Study on the Chain Restaurants Dynamic Negotiation Games of the Optimization of Joint Procurement of Food Materials International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 A Study on the Chain estaurants Dynaic Negotiation aes of the Optiization of Joint Procureent of Food

More information

ESTIMATING LIQUIDITY PREMIA IN THE SPANISH GOVERNMENT SECURITIES MARKET

ESTIMATING LIQUIDITY PREMIA IN THE SPANISH GOVERNMENT SECURITIES MARKET ESTIMATING LIQUIDITY PREMIA IN THE SPANISH GOVERNMENT SECURITIES MARKET Francisco Alonso, Roberto Blanco, Ana del Río and Alicia Sanchis Banco de España Banco de España Servicio de Estudios Docuento de

More information

Equivalent Tapped Delay Line Channel Responses with Reduced Taps

Equivalent Tapped Delay Line Channel Responses with Reduced Taps Equivalent Tapped Delay Line Channel Responses with Reduced Taps Shweta Sagari, Wade Trappe, Larry Greenstein {shsagari, trappe, ljg}@winlab.rutgers.edu WINLAB, Rutgers University, North Brunswick, NJ

More information

The individual neurons are complicated. They have a myriad of parts, subsystems and control mechanisms. They convey information via a host of

The individual neurons are complicated. They have a myriad of parts, subsystems and control mechanisms. They convey information via a host of CHAPTER 4 ARTIFICIAL NEURAL NETWORKS 4. INTRODUCTION Artificial Neural Networks (ANNs) are relatively crude electronic odels based on the neural structure of the brain. The brain learns fro experience.

More information

How To Balance Over Redundant Wireless Sensor Networks Based On Diffluent

How To Balance Over Redundant Wireless Sensor Networks Based On Diffluent Load balancing over redundant wireless sensor networks based on diffluent Abstract Xikui Gao Yan ai Yun Ju School of Control and Coputer Engineering North China Electric ower University 02206 China Received

More information

SUPPORTING YOUR HIPAA COMPLIANCE EFFORTS

SUPPORTING YOUR HIPAA COMPLIANCE EFFORTS WHITE PAPER SUPPORTING YOUR HIPAA COMPLIANCE EFFORTS Quanti Solutions. Advancing HIM through Innovation HEALTHCARE SUPPORTING YOUR HIPAA COMPLIANCE EFFORTS Quanti Solutions. Advancing HIM through Innovation

More information

Airline Yield Management with Overbooking, Cancellations, and No-Shows JANAKIRAM SUBRAMANIAN

Airline Yield Management with Overbooking, Cancellations, and No-Shows JANAKIRAM SUBRAMANIAN Airline Yield Manageent with Overbooking, Cancellations, and No-Shows JANAKIRAM SUBRAMANIAN Integral Developent Corporation, 301 University Avenue, Suite 200, Palo Alto, California 94301 SHALER STIDHAM

More information

COMBINING CRASH RECORDER AND PAIRED COMPARISON TECHNIQUE: INJURY RISK FUNCTIONS IN FRONTAL AND REAR IMPACTS WITH SPECIAL REFERENCE TO NECK INJURIES

COMBINING CRASH RECORDER AND PAIRED COMPARISON TECHNIQUE: INJURY RISK FUNCTIONS IN FRONTAL AND REAR IMPACTS WITH SPECIAL REFERENCE TO NECK INJURIES COMBINING CRASH RECORDER AND AIRED COMARISON TECHNIQUE: INJURY RISK FUNCTIONS IN FRONTAL AND REAR IMACTS WITH SECIAL REFERENCE TO NECK INJURIES Anders Kullgren, Maria Krafft Folksa Research, 66 Stockhol,

More information

Stochastic Online Scheduling on Parallel Machines

Stochastic Online Scheduling on Parallel Machines Stochastic Online Scheduling on Parallel Machines Nicole Megow 1, Marc Uetz 2, and Tark Vredeveld 3 1 Technische Universit at Berlin, Institut f ur Matheatik, Strasse des 17. Juni 136, 10623 Berlin, Gerany

More information