Loading [MathJax]/jax/output/SVG/jax.js
Review Topical Sections

Proposed big data architecture for facial recognition using machine learning

  • With the abundance of raw data generated from various sources including social networks, big data has become essential in acquiring, processing, and analyzing heterogeneous data from multiple sources for real-time applications. In this paper, we propose a big data framework suitable for pre‑processing and classification of image as well as text analytics by employing two key workflows, called big data (BD) pipeline and machine learning (ML) pipeline. Our unique end-to-end workflow integrates data cleansing, data integration, data transformation and data reduction processes, followed by various analytics using suitable machine learning techniques. Further, our model is the first of its kind to augment facial recognition with sentiment analysis in a distributed big data framework. The implementation of our model uses state-of-the-art distributed technologies to ingest, prepare, process and analyze big data for generating actionable data insights by employing relevant ML algorithms such as k-NN, logistic regression and decision tree. In addition, we demonstrate the application of our big data framework to facial recognition system using open sources by developing a prototype as a use case. We also employ sentiment analysis on non-repetitive semi structured public data (text) such as user comments, image tagging, and other information associated with the facial images. We believe our work provides a novel approach to intersect Big Data, ML and Face Recognition and would create new research to alleviate some of the challenges associated with big data processing in real world applications.

    Citation: Suriya Priya R Asaithambi, Sitalakshmi Venkatraman, Ramanathan Venkatraman. Proposed big data architecture for facial recognition using machine learning[J]. AIMS Electronics and Electrical Engineering, 2021, 5(1): 68-92. doi: 10.3934/electreng.2021005

    Related Papers:

    [1] Jian Yang, Yuefen Chen, Zhiqiang Li . Some sufficient conditions for a tree to have its weak Roman domination number be equal to its domination number plus 1. AIMS Mathematics, 2023, 8(8): 17702-17718. doi: 10.3934/math.2023904
    [2] Weisheng Zhao, Ying Li, Ruizhi Lin . The existence of a graph whose vertex set can be partitioned into a fixed number of strong domination-critical vertex-sets. AIMS Mathematics, 2024, 9(1): 1926-1938. doi: 10.3934/math.2024095
    [3] Fu-Tao Hu, Xing Wei Wang, Ning Li . Characterization of trees with Roman bondage number 1. AIMS Mathematics, 2020, 5(6): 6183-6188. doi: 10.3934/math.2020397
    [4] Muhammad Kamran Jamil, Muhammad Imran, Aisha Javed, Roslan Hasni . On the first general Zagreb eccentricity index. AIMS Mathematics, 2021, 6(1): 532-542. doi: 10.3934/math.2021032
    [5] Bana Al Subaiei, Ahlam AlMulhim, Abolape Deborah Akwu . Vertex-edge perfect Roman domination number. AIMS Mathematics, 2023, 8(9): 21472-21483. doi: 10.3934/math.20231094
    [6] Jianhua Tu, Lei Zhang, Junfeng Du, Rongling Lang . Polynomial time recognition of vertices contained in all (or no) maximum dissociation sets of a tree. AIMS Mathematics, 2022, 7(1): 569-578. doi: 10.3934/math.2022036
    [7] Chao Yang, Bing Yao, Zhi-xiang Yin . A new vertex distinguishing total coloring of trees. AIMS Mathematics, 2021, 6(9): 9468-9475. doi: 10.3934/math.2021550
    [8] Abel Cabrera-Martínez, Andrea Conchado Peiró, Juan Manuel Rueda-Vázquez . Further results on the total Italian domination number of trees. AIMS Mathematics, 2023, 8(5): 10654-10664. doi: 10.3934/math.2023540
    [9] Saeed Kosari, Yongsheng Rao, Zehui Shao, Jafar Amjadi, Rana Khoeilar . Complexity of signed total k-Roman domination problem in graphs. AIMS Mathematics, 2021, 6(1): 952-961. doi: 10.3934/math.2021057
    [10] Tariq A. Alraqad, Igor Ž. Milovanović, Hicham Saber, Akbar Ali, Jaya P. Mazorodze, Adel A. Attiya . Minimum atom-bond sum-connectivity index of trees with a fixed order and/or number of pendent vertices. AIMS Mathematics, 2024, 9(2): 3707-3721. doi: 10.3934/math.2024182
  • With the abundance of raw data generated from various sources including social networks, big data has become essential in acquiring, processing, and analyzing heterogeneous data from multiple sources for real-time applications. In this paper, we propose a big data framework suitable for pre‑processing and classification of image as well as text analytics by employing two key workflows, called big data (BD) pipeline and machine learning (ML) pipeline. Our unique end-to-end workflow integrates data cleansing, data integration, data transformation and data reduction processes, followed by various analytics using suitable machine learning techniques. Further, our model is the first of its kind to augment facial recognition with sentiment analysis in a distributed big data framework. The implementation of our model uses state-of-the-art distributed technologies to ingest, prepare, process and analyze big data for generating actionable data insights by employing relevant ML algorithms such as k-NN, logistic regression and decision tree. In addition, we demonstrate the application of our big data framework to facial recognition system using open sources by developing a prototype as a use case. We also employ sentiment analysis on non-repetitive semi structured public data (text) such as user comments, image tagging, and other information associated with the facial images. We believe our work provides a novel approach to intersect Big Data, ML and Face Recognition and would create new research to alleviate some of the challenges associated with big data processing in real world applications.



    All graphs considered in this paper are finite, undirected and simple. Let G be a graph with vertex-set V(G) and edge-set E(G). A subset D of V(G) is called a dominating set of G if every vertex of G is either in D or adjacent to a vertex of D. The domination number γ(G) is the cardinality of a minimum dominating set of G.

    The domination in graphs is so classic that it has been widely studied in networks theory, while the decision problem for the domination number of a general graph was proved to be NP-complete [8]. On the study of domination, there are three early textbooks compiled by Haynes et al. [13,14] and Henning et al. [17]. Recently, Haynes, Hedetniemi and Henning [11,12] edited two new books on this field once again.

    Definition 1.1. [24] (1) A vertex vV(G) is called γ-fixed if v belongs to every minimum dominating set of G. (2) A vertex vV(G) is called γ-bad if v does not belong to any minimum dominating set of G. (For simplicity, we abbreviate "γ-fixed" and "γ-bad" to "fixed" and "bad" respectively in this paper.)

    Definition 1.2. A vertex vV(G) is called critical if γ(Gv)<γ(G). In particular, we agree that the single vertex of a trivial graph is critical.

    Remark for Definition 1.2: It is easy to see that γ(Gv)<γ(G)γ(Gv)γ(G)1γ(Gv)=γ(G)1, where γ(Gv)γ(G)1γ(Gv)=γ(G)1 holds because if not so, then γ(Gv)γ(G)2, and thus G would have a dominating set with cardinality ((γ(G)2)+|v|)<γ(G), contradicting the minimality of γ(G).

    The terms of fixed and bad vertices of graphs were introduced by Samodivkin [24], which can help us to research the constructions of minimum dominating sets of a graph better and shorten the processes of our proofs [7,22,24,25]. (In [28], fixed and bad vertices of a graph are also called universal and idle vertices, respectively.) The notions of domination critical, which include vertex-critical [4] and edge-critical [3,27], are very important to domination of graphs. But in this paper, we are not going to discuss the topic of edge-critical.

    Definition 1.3. A graph G is called vertex-critical if every vertex of G is critical.

    There are lots of nice properties on domination vertex-critical graphs [1,2,19,29,30]. Many of them possess symmetry, and even vertex-transitivity. For example, the graph C4C4 obtained by identifying two vertices of two cycles of orders four, the Harary graph H3,8 and the circulant graph C121,5 (See Figure 1.).

    Figure 1.  Three examples of vertex-critical graph.

    Trees is a kind of basic graph class often applied to algorithm design. There are a good few results on the study of the constructions for special trees, such as trees with equal domination and total domination numbers [7], trees with equal domination and restrained domination numbers [6], trees with equal total domination and disjunctive domination numbers [18], trees with equal independent domination and weak domination numbers [10], trees with a minimum vertex cover also being a minimum total dominating set [5], trees with two disjoint minimum independent dominating sets [15], trees with the paired domination number being twice the matching number [26], trees without fixed vertices [31], trees without fixed vertices and critical vertices [23,16], trees with unique minimum dominating sets [9,34], trees with equal Roman {2}-domination and Roman domination numbers [21], and trees with total Roman domination number being equal to the sum of domination number and semitotal domination number [20].

    Naturally, there are two such questions: Can we exhaustively characterize vertex-critical graphs, as well as graphs without critical vertices? It seems not easy to solve these two questions. Therefore, in this paper, we study on the latter one and focus on the graph class-trees. Via defining 3 operations of graphs, we get a constructive characterization of trees without critical vertices.

    For any u,vV(G), denote by dG(u,v) the distance from u to v in G as well as dG(v), NG(v), NG[v] and N2G(v) the degree, open neighborhood, closed neighborhood and 2-open neighborhood of vertex v in G respectively, where the 2-open neighborhood of vertex v in G is defined as N2G(v)={xV(G)d(x,v)=2}. For any XV(G), let G[X] denote the subgraph of G induced by X.

    Denote by MDS_(G) the set composed of all the minimum dominating sets of G. That is, MDS_(G)={DD is a minimum dominating set of G}. A vertex of degree one (resp. degree zero) in G is called an end-vertex (resp. isolated vertex) of G. Let g be a cut-vertex of G. If a component P of Gg is a path and g is adjacent to an end-vertex of P in G, then we call P as a pendant path of G and say that g and P are linked with each other. A pendant path of G with order l (l1) is called an l-pendant path of G. Let P2k+1=v1v2v2k+1 be a path of order 2k+1. Then vk+1 is the center of P2k+1.

    Let r be a vertex, l and m be two non-negative integers with l+m1. Let P13P23Pl3P3 with Pi3=viuiwi, i=1,2,,l, and P12P22Pm2P2 with Pj2=xjyj, j=1,2,,m. For every 1il, link r and ui by an edge. For every 1jm, link r and xj by an edge. Denote the resulting graph as Rl,m and call r as the root of Rl,m (See Figure 2).

    Figure 2.  The graph Rl,m.

    Now, we give three observations and some known lemmas, which will support our proofs in the next section.

    Observation 2.1. Let G be a graph. If G1 and G2 are vertex-induced subgraphs of G such that V(G)=V(G1)V(G2), then γ(G)γ(G1)+γ(G2) with the equality holding when G1 and G2 are two components of G.

    Observation 2.2. Let uV(G). If u is adjacent to two end-vertices v and w in G, then v and w are bad in G, and u is fixed in G.

    Observation 2.3. Let u be an end-vertex of G with NG(u)={v}. Then u is non-fixed and v is non-bad in G, and |{u,v}D|=1 for any DMDS_(G).

    Lemma 2.4. [3] For any nontrivial tree T and any vV(T), v is a fixed vertex of T if and only if γ(Tv)>γ(T).

    Lemma 2.5. [32] Let G be a graph with minimum degree at least one. If x is a bad or fixed vertex of G, then all the elements of NG[x] are non-critical vertices of G.

    Lemma 2.6. [33] Let G be a graph.

    (a) If x is a non-fixed vertex of G, then γ(Gx)γ(G).

    (b) If x is a bad vertex of G, then γ(Gx)=γ(G).

    (c) If x is a non-fixed and non-critical vertex of G, then γ(Gx)=γ(G).

    Lemma 2.7. [34] Let T be a tree containing only one vertex u of degree at least 3. Then u is linked with |NT(u)| pendant paths in T.

    Lemma 2.8. [34] Let T be a tree with at least two vertices of degree at least 3 and let dT(u,v)=max{dT(x,y)| both x and y are vertices of degree at least 3 in T}. Then u is linked with |NT(u)|1 pendant paths in T.

    Lemma 2.9. [34] Let G0 be a graph without any isolated vertices and possessing a fixed vertex. If G is a graph obtained via linking a fixed vertex of G0 and the single vertex of P1 by an edge, then γ(G)=γ(G0).

    Lemma 2.10. [34] Let G0 be a graph without any isolated vertices and possessing a fixed vertex. If G is a graph obtained via linking a fixed vertex of G0 and the center of P3 by an edge, then γ(G)=γ(G0)+1.

    We now ask a question: Is there a nontrivial tree only containing critical vertices? Unluckily, the answer to this question is no (See Lemma 3.1).

    Lemma 3.1. If dG(u)=1 and vNG(u)N2G(u), then v is a non-critical vertex of G.

    Proof. Suppose to the contrary that v is a critical vertex of G. If vNG(u), let D1MDS_(Gv). Then by Definition 1.2, we have |D1|=γ(G)1. Since u is an isolated vertex of Gv, it follows that u is fixed in Gv. Now, if we let D2=(D1{u}){v}, then D2 is a dominating set of G with |D2|=|D1|=γ(G)1, a contradiction.

    If vN2G(u), then |D|=γ(G)1 for any DMDS_(Gv). Let NG(u)={w}. Since u is still an end-vertex of Gv, it follows from Observation 2.3 that w is a non-bad vertex of Gv. Let D1MDS_(Gv) with wD1. On one hand, we have |D1|=γ(G)1. But on the other hand, D1 is also a dominating set of G, which implies that |D1|γ(G), a contradiction.

    Lemma 3.1 tells us that if G is nontrivial and has an end-vertex, then G must have a non-critical vertex. Therefore, a tree is vertex-critical if and only if it is trivial.

    In this subsection, via several operations of graphs, we can get large graphs without critical vertices from small graphs without critical vertices step by step. In particular, these processes of operations are reversible for trees. (Here, large graph represents graph with large order while small graph represents graph with small order.) For a graph G0, we define the following three operations.

    Operation i. Link a fixed vertex of G0 and the single vertex of P1 by an edge. Denote the resulting graph by G0P1. (Refer to Figure 3 (i)).

    Figure 3.  Examples of G0P1, G0P3 and G0Rl,m.

    Operation ii. Link a fixed vertex of G0 and the center of P3 by and edge. Denote the resulting graph by G0P3. (Refer to Figure 3 (ii)).

    Operation iii. Link an arbitrary vertex of G0 and the root of Rl,m by and edge. Denote the resulting graph by G0Rl,m. (Refer to Figure 3 (iii)).

    Remark. In fact, the resulting graph may be not unique. So, "G=G0P1" means that "G is obtained from G0 by Operation i".

    Lemma 3.2. Let G0 be a graph without any isolated vertices and possessing a fixed vertex, and let G=G0P1. Then all the vertices of G0 are non-critical if and only if all the vertices of G are non-critical.

    Proof. () Suppose that V(P1)={v1} and gv1E(G). From Lemma 2.9, we have γ(G)=γ(G0)=γ(Gv1). So by Definition 1.2, v1 is a non-critical vertex of G. Also, we have g is a non-critical vertex of G by Lemma 3.1. It remains to prove that x is a non-critical vertex of G for every xV(Gv1g)V(G0). Since all the vertices of G0 are non-critical, we have γ(G0x)γ(G0). Since dGx(v1)=1, we have v1 is a non-fixed vertex of Gx by Observation 2.3. So there exists DxMDS_(Gx) such that v1Dx, and then gDx. Thus DxV(G0x) is a dominating set of G0x. Hence γ(Gx)=|Dx|=|DxV(G0x)|γ(G0x)γ(G0)=γ(G), which implies that x is a non-critical vertex of G. The necessity follows.

    () Assume to the contrary that G0 has a critical vertex y0. Since all the vertices of G are non-critical, we have γ(Gy0)γ(G). Let D0MDS_(G0y0). Then |D0{y0}|=γ(G0y0)+1=γ(G0), which implies that D0{y0}MDS_(G0). By the definition of Operation i, g is a fixed vertex of G0. So we have gD0{y0}, and g is a non-critical vertex of G0 by Lemma 2.5. Thus gy0, and therefore gD0, which implies that D0 is a dominating set of Gy0. Hence γ(G0y0)=|D0|γ(Gy0)γ(G)=γ(G0), which contradicts the assumption that y0 is a critical vertex of G0. The sufficiency follows.

    Lemma 3.3. Let G0 and W be two graphs. Let G be a graph obtained via linking an arbitrary vertex of G0 and an arbitrary vertex of W by an edge. If γ(G)=γ(G0)+γ(W) and all the vertices of G are non-critical, then all the vertices of G0 are non-critical.

    Proof. For any yV(G0), since all the vertices of G are non-critical, it follows that γ(Gy)γ(G). By Observation 2.1, we have γ(G0y)+γ(W)γ(Gy)γ(G)=γ(G0)+γ(W). Thus γ(G0y)γ(G0), and so y is a non-critical vertex of G0. The lemma follows.

    Lemma 3.4. Let G0 be a graph without isolated vertices and possessing a fixed vertex, and G=G0P3.

    (a) If all the vertices of G are non-critical, then all the vertices of G0 are non-critical.

    (b) When G0 is a tree, (in order to avoid confusion, ) we rewrite T0=G0 and T=G. If all the vertices of T0 are non-critical, then all the vertices of T are non-critical.

    Proof. (a) Suppose that P3=v1v2v3 and E(G)E(G0)E(P3)={gv2}. Then g is fixed in G0. From Lemma 2.10, we get γ(G)=γ(G0)+1. Item (a) follows by Lemma 3.3.

    (b) Firstly, by Observation 2.2, v2 is a fixed vertex of T. So by Lemma 2.5, v1,v2, v3 and g are non-critical vertices of T.

    Secondly, we need to show that x is a non-critical vertex of T for every xV(Tg)V(P3). That is, to prove γ(Tx)γ(T). Since T0 has no critical vertices, we have γ(T0x)γ(T0). Let DxMDS_(Tx). If gDx, then DxV(T0x) is a dominating set of T0x, and so γ(Tx)=|Dx|=|DxV(T0x)|+|{v2}|γ(T0x)+1γ(T0)+1=γ(T). If gDx, then (DxV(T0x)){x} is a dominating set of T0g. By Lemma 2.4, we have |(DxV(T0x)){x}|γ(T0g)γ(T0)+1, which implies that |DxV(T0x)|γ(T0). So γ(Tx)=|Dx|=|DxV(T0x)|+|{v2}|γ(T0)+1=γ(T). Item (b) follows.

    Note. In Lemma 3.4 (b), we restrict G0 to be a tree because if G0 is a general graph, then the result maybe not true. (See the following Example 3.5).

    Example 3.5. Define G0 and G as shown in Figure 4. Then G=G0P3. It is not hard to check that γ(G0)=2, h1 and g are fixed vertices of G0, as well as w1,u1,u2,u3,u4 and h2 are bad vertices of G0. By Lemma 2.5, G0 has no critical vertices. However, since γ(G)=γ(G0)+1=3, we can see that {h2,v2}MDS_(Gw1), which implies that w1 is a critical vertex of G.

    Figure 4.  Sketch of Example 3.5.

    Lemma 3.6. Let G0 be a graph and G=G0Rl,m. Then

    (a)γ(G)=γ(G0)+(l+m);

    (b) all the vertices of G0 are non-critical if and only if all the vertices of G are non-critical.

    Proof. Suppose that E(G)E(G0)E(Rl,m)={gr}. Set U={u1,u2,,ul}, X={x1,x2,,xm}, Y={y1,y2,,ym} and Z=XY.

    (a) We can easily see that γ(Rl,m)=l+m. So γ(G)γ(G0)+(l+m). It remains to prove γ(G)γ(G0)+(l+m). Let DMDS_(G). If DV(G0) can dominate g in G, then DV(G0) is a dominating set of G0, and so |D||DV(G0)|+|U|+|DZ|γ(G0)+(l+m); if not, then rD and (DV(G0)){g} is a dominating set of G0, and so |D|=|DV(G0)|+|{r}|+|U|+|DZ|=|(DV(G0)){g}|+(l+m)γ(G0)+(l+m).

    (b) () The sufficiency follows immediately by Item (a) and Lemma 3.3.

    () We claim that r is a bad vertex of G. Otherwise, let DrMDS_(G) with rDr. If gDr, then Dr{r} is a dominating set of G, contradicting the minimality of |Dr|. So we have gDr, and then DrV(G0) is a dominating set of G0g. Thus γ(G0g)|DrV(G0)|=|Dr||U||DrZ||{r}|=γ(G)(l+m)1=γ(G0)1, which implies that g is a critical vertex of G0, contradicting the known condition that G0 has no critical vertices.

    Now firstly, for every 1il and every 1jm, we have ui,vi,wi,r,g and xj are non-critical vertices of G by Lemma 2.5.

    Secondly, if there exists some 1jm such that γ(Gyj)=γ(G)1, we can let DMDS_(Gyj) with rD by Observation 2.3. But then D{yj}MDS_(G) with rD{yj}, contradicting the claim that r is bad in G. So for every 1jm, we have γ(Gyj)γ(G), which implies that yj is non-critical in G.

    Finally, it remains to show that γ(Gx)γ(G) for every xV(Gg)V(Rl,m)V(G0). Since all the vertices of G0 are non-critical, we have γ(G0x)γ(G0). Let DxMDS_(Gx). If DxV(G0x) can dominate g, then DxV(G0x) is a dominating set of G0x, and so γ(Gx)=|Dx||DxV(G0x)|+|U|+|DxZ|γ(G0x)+(l+m)γ(G0)+(l+m)=γ(G); if not, then we have rDx and (DxV(G0x)){g} is a dominating set of G0x, and so γ(Gx)=|Dx|=|DxV(G0x)|+|{r}|+|U|+|DxZ|=|(DxV(G0x)){g}|+(l+m)γ(G0x)+(l+m)γ(G0)+(l+m)=γ(G). The necessity follows.

    Since it is hard to obtain a constructive characterization of graphs without critical vertices, we only solve this problem partly by restricting the graph class to be trees in this subsection.

    Theorem 3.7. A nontrivial tree T has no critical vertices if and only if T can be obtained from P2 or P3 by a finite sequence of Operations iiii.

    Proof. Let T be the set of graphs obtained from P2 or P3 by a finite sequence of Operations i–iii. It suffices to prove that T has no critical vertices if and only if TT.

    () Assume that T is obtained by doing n times Operations i, ii and iii. We will prove that all the vertices of T are non-critical by induction on n. When n=0, we have T=P2 or T=P3, and the result is true clearly. Suppose that the result is true when n=k(k0). Then from Lemmas 3.2, 3.4 (b), and 3.6 (b), we know that the result is also true when n=k+1. By the induction principle, the sufficiency follows.

    () We are going to prove the necessity by induction on |V(T)|. When |V(T)|=2 or 3, the result is true clearly. Suppose that the result is true when |V(T)|<k(k4). We consider the case when |V(T)|=k below.

    Case 1. T has a pendant path of order at least 3.

    Let P3 be a 3-pendant path of T and T0=TV(P3). Note that P3R0,1. So T=T0R0,1. By Lemma 3.6 (b), all the vertices of T0 are non-critical. Since |V(T0)|<|V(T)|=k, we have T0T by the induction hypothesis. Hence T=T0R0,1T.

    Case 2. T has a vertex u adjacent to an end-vertex w in T and u is fixed in Tw.

    Let T0=Tw. Then T=T0P1. By Lemma 3.2, T0 has no critical vertices. Since |V(T0)|=k1<k, we have T0T. Hence T=T0P1T.

    Case 3. T has a vertex u of degree 3 which is adjacent to two end-vertices v,w of T and a fixed vertex g of T{v,u,w} in T.

    Let T0=T{v,u,w}. Then T=T0P3. By Lemma 3.4 (a), T0 has no critical vertices. So T0T, and hence TT.

    Case 4. T has a vertex u of degree at least 3 linked with |NT(u)|1 2-pendant paths.

    Let P12,P22,,Pm2 be the 2-pendant paths linked with u in T, where m=|NT(u)|1. Then T[{u}mj=1V(Pj2)]R0,m. Let T0=T({u}mj=1V(Pj2)). Then T=T0R0,m. As a consenquence, we have TT.

    Case 5. All of Cases 1–4 do not occur.

    Since |V(T)|4 and Case 1 does not occur, T is not a path. So T has at least one vertex of degree at least 3.

    Claim 5.1. There does not exist a vertex u adjacent to three end-vertices in T.

    Suppose not. Let v1 be an end-vertex which is adjacent to u in T and let T0=Tv1. By Observation 2.2, u is a fixed vertex of T0, which is contrary to the supposition that Case 2 does not occur.

    Claim 5.2. There does not exist a vertex u linked with one 1-pendant path P1 and one 2-pendant path P2 in T.

    Suppose not. Let V(P2)={x,y} with uxE(T). By Observation 2.3, there exists DuMDS_(T) such that uDu. Let Du=(Du{x,y}){y}. Then Du{y} is a dominating set of Ty, which implies that y is a critical vertex of T, contradicting the known condition that T has no critical vertices.

    Claim 5.3. T has at least two vertices of degree at least 3.

    Suppose, to the contrary, that T has only one vertex c with dT(c)3. By Lemma 2.7, c is linked with |NT(c)| pendant paths in T. Since Case 1 does not occur, it follows from Claims 5.2 and 5.1 that all of these |NT(c)| pendant paths are 2-pendant paths. But this contradicts the supposition that Case 4 does not occur.

    Claim 5.4. If u and v are two vertices of degree at least 3 in T such that dT(u,v)= max {dT(x,y)|both x and y are vertices of degrees at least 3 in T}, then |NT(u)|=3 and u is adjacent to 2 end-vertices in T.

    By Lemma 2.8, u is linked with |NT(u)|1 pendant paths in T. Since Cases 1 and 4 does not occur, it follows from Claims 5.2 and 5.1 that |NT(u)|=3 and u is adjacent to 2 end-vertices in T.

    Now, let u and v be two vertices of T satisfying the supposition of Claim 5.4. Suppose that v1 and w1 are two end-vertices which are adjacent to u in T and NT(u){v1,w1}={r}. By Claim 5.4, the equivalent status of v and u, and Observation 2.2, we get that v is a fixed vertex of T{u,v1,w1}. Since Case 3 does not occur, we have rv.

    Let Tv be the component of Tr such that vV(Tv), {z}=NT(r)V(Tv), NT(r){z}={u1,u2,,uq} (where u1=u) and Tu1,Tu2,,Tuq be the components of Tr such that uiV(Tui) for every 1iq. Furthermore, we may suppose without loss of generality that Tu1,Tu2,,Tul are not pendant paths of T as well as Tul+1,Tul+2,,Tul+m are pendant paths of T, where 1lq and l+m=q.

    Claim 5.5. Tui is a 2-pendant path of T for every l+1il+m.

    Suppose, to the contrary, that Tui is a pendant path of T with |V(Tui)|2 for some l+1il+m. By Lemma 3.1, r is a non-critical vertex of TV(Tui). Since Case 1 does not occur, we have |V(Tui)|=1. Let V(Tui)={w} and T0=T{u,v1,w1}. By Observations 2.2 and 2.3, one may let DrMDS_(T) with u,rDr. We claim that DrV(T0)MDS_(T0). Otherwise there exists D0MDS_(T0) such that |D0|<|DrV(T0)|. But then {u}D0 would be a dominating set of T with |{u}D0|<|{u}(DuV(T0))|=|Dr|, contradicting the minimality of |Dr|. Thus γ(T)=|Dr|=|{u}|+|DrV(T0)|=1+γ(T0). Since Case 3 does not occur, r is not a fixed vertex of T0. Let ˆDr0MDS_(T0) with rˆDr0. Then wˆDr0 and {u}ˆDr0MDS_(T). Since ({u}ˆDr0){w} is a dominating set of Tw, it follows that w is a critical vertex of T, a contradiction.

    Claim 5.6. For every 2il, ui is the unique vertex of V(Tui) satisfying dT(ui)3.

    Firstly, since Tui is not a pendant path of T, Tui has a vertex with degree at least 3 in T. Secondly, we claim that for every hV(Tui){ui}, h is not a vertex of V(Tui) with degree at least 3 in T. Otherwise, we have dT(h,v)=dT(h,ui)+dT(ui,v)>dT(ui,v)=1+dT(r,v)=dT(u,v), contradicting the selection of u and v. From these two observations, Claim 5.6 follows.

    Since dT(ui,v)=dT(u1,v), we have |NT(ui)|=3 and ui is adjacent to 2 end-vertices in T for every 2il by Claim 5.4, which implies that

    T[{r}li=1V(Tui)mj=1V(Tul+j)]Rl,m.

    (See Figure 5.) Let T0=T({r}li=1V(Tui)mj=1V(Tul+j)). Then T=T0Rl,m. By Lemma 3.6 (b), T0 has no critical vertices. Thus we have T0T by the induction hypothesis, and so TT.

    Figure 5.  The sketch of Case 5.

    In conclusion, the result is true when |V(T)|=k. The necessity follows.

    We think that it is quite difficult to give a construction for graphs without critical vertices. For further studies, ones may consider to characterize unicyclic graphs without critical vertices, or graphs with domination number 3 and without critical vertices.

    1. National Natural Science Foundation of China (no. 12061047);

    2. Natural Science Foundation of Jiangxi Province (no. 20192BAB211002);

    3. Undergraduate Innovation Training Project of Hubei Province (no. 202111072011);

    4. Foundation of Cultivation of Scientific Institutions of Jianghan University (no. 06210033).

    The authors declared that they have no conflicts of interest to this work.



    [1] Chen M, Mao S, Liu Y (2014) Big data: A survey. Mobile networks and applications 19: 171-209. doi: 10.1007/s11036-013-0489-0
    [2] McAfee A, Brynjolfsson E, Davenport TH, et al. (2012) Big data: the management revolution. Harvard business review 90: 60-68.
    [3] Venkatraman R, Venkatraman S (2019) Big Data Infrastructure, Data Visualisation and Challenges. Proceedings of the 3rd International Conference on Big Data and Internet of Things, 13-17.
    [4] Labrinidis A, Jagadish HV (2012) Challenges and opportunities with big data. Proceedings of the VLDB Endowment 5: 2032-2033. doi: 10.14778/2367502.2367572
    [5] Venkatraman S, Venkatraman R (2019) Big data security challenges and strategies. AIMS MATHEMATICS 4: 860-879. doi: 10.3934/math.2019.3.860
    [6] Masi I, Wu Y, Hassner T, et al. (2018) Deep face recognition: A survey. 2018 31st SIBGRAPI conference on graphics, patterns and images (SIBGRAPI), 471-478.
    [7] Singh A, Bhadani R (2020) Mobile Deep Learning with TensorFlow Lite, ML Kit and Flutter. Packt Publishing.
    [8] Zhu Y, Jiang Y (2020) Optimization of face recognition algorithm based on deep learning multi feature fusion driven by big data. Image Vision Comput 104: 104023. doi: 10.1016/j.imavis.2020.104023
    [9] Reddy KS, Krishna VV, Kumar VV (2016) A Method for Facial Recognition Based On Local Features. International Journal of Mathematics and Computation 27: 98-109.
    [10] Qateef JS, Kazm AA (2016) Facial expression recognition via mapreduce assisted k-nearest neighbor algorithm. International Journal of Computer Science and Information Security 14: 170.
    [11] Sirovich L, Kirby M (1987) Low-dimensional procedure for the characterization of human faces. Josa a 4: 519-524. doi: 10.1364/JOSAA.4.000519
    [12] Turk MA, Pentland AP (1991) Face recognition using eigenfaces. Proceedings. 1991 IEEE computer society conference on computer vision and pattern recognition, 586-587. IEEE Computer Society.
    [13] Bruce V, Young A (1986) Understanding face recognition. British journal of psychology 77: 305-327. doi: 10.1111/j.2044-8295.1986.tb02199.x
    [14] Parkhi OM, Vedaldi A, Zisserman A (2015) Deep face recognition.
    [15] He X, Yan S, Hu Y, et al. (2005) Face recognition using Laplacianfaces. IEEE T Pattern Anal 27: 328-340. doi: 10.1109/TPAMI.2005.55
    [16] Deng J, Guo J, Xue N, et al. (2019) Arcface: Additive angular margin loss for deep face recognition. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 4690-4699.
    [17] Zhou E, Cao Z, Yin Q (2015) Naive-deep face recognition: Touching the limit of LFW benchmark or not? arXiv preprint arXiv: 150104690.
    [18] Wang H, Wang Y, Zhou Z, et al. (2018) Cosface: Large margin cosine loss for deep face recognition. Proceedings of the IEEE conference on computer vision and pattern recognition, 5265-5274.
    [19] Wen Y, Zhang K, Li Z, et al. (2016) A discriminative feature learning approach for deep face recognition. European conference on computer vision, 499-515.
    [20] Deng J, Zhou Y, Zafeiriou S (2017) Marginal loss for deep face recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 60-68.
    [21] Ding H, Zhou SK, Chellappa R (2017) Facenet2expnet: Regularizing a deep face recognition net for expression recognition. 2017 12th IEEE International Conference on Automatic Face & Gesture Recognition (FG 2017), 118-126.
    [22] Wang F, Chen L, Li C, et al. (2018) The devil of face recognition is in the noise. Proceedings of the European Conference on Computer Vision (ECCV), 765-780.
    [23] Benhlima L (2018) Big data management for healthcare systems: architecture, requirements, and implementation. Advances in bioinformatics 2018.
    [24] Asaithambi SPR, Venkatraman R, Venkatraman S (2020) MOBDA: Microservice-Oriented Big Data Architecture for Smart City Transport Systems. Big Data and Cognitive Computing 4: 17. doi: 10.3390/bdcc4030017
    [25] Costa C, Santos MY (2016) BASIS: A big data architecture for smart cities. 2016 SAI Computing Conference (SAI), 1247-1256.
    [26] He X, Wang K, Huang H, et al. (2018) QoE-driven big data architecture for smart city. IEEE Commun Mag 56: 88-93. doi: 10.1109/MCOM.2018.1700231
    [27] Lopez D, Manogaran G (2016) Big data architecture for climate change and disease dynamics. The human element of big data: issues, analytics, and performance, 301-331.
    [28] Dean J, Ghemawat S (2008) MapReduce: simplified data processing on large clusters. Communications of the ACM 51: 107-113. doi: 10.1145/1327452.1327492
    [29] Peralta D, Del Río S, Ramírez-Gallego S, et al. (2015) Evolutionary feature selection for big data classification: A mapreduce approach. Math Probl Eng 2015.
    [30] Gao W, Zhao X, Gao Z, et al. (2019) 3D Face Reconstruction From Volumes of Videos Using a Mapreduce Framework. IEEE Access 7: 165559-165570. doi: 10.1109/ACCESS.2019.2938671
    [31] Mahmoud SM, Habeeb RS (2019) Analysis of Large Set of Images Using MapReduce Framework. International Journal of Modern Education and Computer Science 11: 47. doi: 10.5815/ijmecs.2019.12.05
    [32] Apache Spark™. A unified analytics engine for large-scale data processing.
    [33] Hazarika AV, Ram GJSR, Jain E (2017) Performance comparision of Hadoop and spark engine. 2017 International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud)(I-SMAC), 671-674.
    [34] Zaharia M, Chowdhury M, Das T, et al. (2012) Resilient distributed datasets: A fault-tolerant abstraction for in-memory cluster computing. 9th {USENIX} Symposium on Networked Systems Design and Implementation ({NSDI} 12), 15-28.
    [35] Luengo J, García S, Herrera F (2012) On the choice of the best imputation methods for missing values considering three groups of classification methods. Knowl Inf Syst 32: 77-108. doi: 10.1007/s10115-011-0424-2
    [36] Batista GE, Monard MC (2003) An analysis of four missing data treatment methods for supervised learning. Appl Artif Intell 17: 519-533. doi: 10.1080/713827181
    [37] Wilson DL (1972) Asymptotic properties of nearest neighbor rules using edited data. IEEE T Syst Man Cy B, 408-421.
    [38] Sánchez JS, Barandela R, Marqués AI, et al. (2003) Analysis of new techniques to obtain quality training sets. Pattern Recogn Lett 24: 1015-1022. doi: 10.1016/S0167-8655(02)00225-8
    [39] Garcia S, Derrac J, Cano J, et al. (2012) Prototype selection for nearest neighbor classification: Taxonomy and empirical study. IEEE T Pattern Anal 34: 417-435. doi: 10.1109/TPAMI.2011.142
    [40] Triguero I, Derrac J, Garcia S, et al. (2011) A taxonomy and experimental study on prototype generation for nearest neighbor classification. IEEE T Syst Man Cy C 42: 86-100. doi: 10.1109/TSMCC.2010.2103939
    [41] García-Gil D, Luengo J, García S, et al. (2019) Enabling smart data: noise filtering in big data classification. Inform Sciences 479: 135-152. doi: 10.1016/j.ins.2018.12.002
    [42] Xue B, Zhang M, Browne WN, et al. (2015) A survey on evolutionary computation approaches to feature selection. IEEE T Evolut Comput 20: 606-626. doi: 10.1109/TEVC.2015.2504420
    [43] Navot A, Shpigelman L, Tishby N, et al. (2005) Nearest neighbor based feature selection for regression and its application to neural activity. Advances in neural information processing systems 18: 996-1002.
    [44] Ramírez-Gallego S, García S, Xiong N, et al. (2018) BELIEF: A distance-based redundancy-proof feature selection method for Big Data. arXiv preprint arXiv: 180405774.
    [45] Triguero I, Peralta D, Bacardit J, et al. (2015) MRPR: a MapReduce solution for prototype reduction in big data classification. Neurocomputing 150: 331-345. doi: 10.1016/j.neucom.2014.04.078
    [46] García-Gil D, Ramírez-Gallego S, García S, et al. (2018) On the Use of Random Discretization and Dimensionality Reduction in Ensembles for Big Data. International Conference on Hybrid Artificial Intelligence Systems, 15-26.
    [47] Triguero I, Galar M, Vluymans S, et al. (2015) Evolutionary undersampling for imbalanced big data classification. 2015 IEEE Congress on Evolutionary Computation (CEC), 715-722.
    [48] Triguero I, Galar M, Merino D, et al. (2016) Evolutionary undersampling for extremely imbalanced big data classification under apache spark. 2016 IEEE Congress on Evolutionary Computation (CEC), 640-647.
    [49] Ramírez-Gallego S, García S, Benítez JM, et al. (2018) A distributed evolutionary multivariate discretizer for big data processing on apache spark. Swarm Evol Comput 38: 240-250. doi: 10.1016/j.swevo.2017.08.005
    [50] Maillo J, Triguero I, Herrera F (2015) A mapreduce-based k-nearest neighbor approach for big data classification. 2015 IEEE Trustcom/BigDataSE/ISPA 2: 167-172. doi: 10.1109/Trustcom.2015.577
    [51] Maillo J, Ramírez S, Triguero I, et al. (2017) kNN-IS: An Iterative Spark-based design of the k-Nearest Neighbors classifier for big data. Knowl-Based Syst 117: 3-15. doi: 10.1016/j.knosys.2016.06.012
    [52] Deng Z, Zhu X, Cheng D, et al. (2016) Efficient kNN classification algorithm for big data. Neurocomputing 195: 143-148. doi: 10.1016/j.neucom.2015.08.112
    [53] Gallego A-J, Calvo-Zaragoza J, Valero-Mas JJ, et al. (2018) Clustering-based k-nearest neighbor classification for large-scale data with neural codes representation. Pattern Recogn 74: 531-543. doi: 10.1016/j.patcog.2017.09.038
    [54] Wang F, Wang Q, Nie F, et al. (2018) Efficient tree classifiers for large scale datasets. Neurocomputing 284: 70-79. doi: 10.1016/j.neucom.2017.12.061
  • This article has been cited by:

    1. Weisheng Zhao, Ruizhi Lin, Junqing Cai, On construction for trees making the equality hold in Vizing's conjecture, 2022, 101, 0364-9024, 397, 10.1002/jgt.22833
    2. Weisheng Zhao, Ying Li, Ruizhi Lin, The existence of a graph whose vertex set can be partitioned into a fixed number of strong domination-critical vertex-sets, 2023, 9, 2473-6988, 1926, 10.3934/math.2024095
  • Reader Comments
  • © 2021 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(6625) PDF downloads(403) Cited by(8)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog