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

A review of the application of machine learning in adult obesity studies


  • Received: 24 January 2022 Revised: 30 March 2022 Accepted: 31 March 2022 Published: 31 March 2022
  • In obesity studies, several researchers have been applying machine learning tools to identify factors affecting human body weight. However, a proper review of strength, limitations and evaluation metrics of machine learning algorithms in obesity is lacking. This study reviews the status of application of machine learning algorithms in obesity studies and to identify strength and weaknesses of these methods. A scoping review of paper focusing on obesity was conducted. PubMed and Scopus databases were searched for the application of machine learning in obesity using different keywords. Only English papers in adult obesity between 2014 and 2019 were included. Also, only papers that focused on controllable factors (e.g., nutrition intake, dietary pattern and/or physical activity) were reviewed in depth. Papers on genetic or childhood obesity were excluded. Twenty reviewed papers used machine learning algorithms to identify the relationship between the contributing factors and obesity. Regression algorithms were widely applied. Other algorithms such as neural network, random forest and deep learning were less exploited. Limitations regarding data priori assumptions, overfitting and hyperparameter optimization were discussed. Performance metrics and validation techniques were identified. Machine learning applications are positively impacting obesity research. The nature and objective of a study and available data are key factors to consider in selecting the appropriate algorithms. The future research direction is to further explore and take advantage of the modern methods, i.e., neural network and deep learning, in obesity studies.

    Citation: Mohammad Alkhalaf, Ping Yu, Jun Shen, Chao Deng. A review of the application of machine learning in adult obesity studies[J]. Applied Computing and Intelligence, 2022, 2(1): 32-48. doi: 10.3934/aci.2022002

    Related Papers:

    [1] Wei Qi . The polycyclic codes over the finite field Fq. AIMS Mathematics, 2024, 9(11): 29707-29717. doi: 10.3934/math.20241439
    [2] Turki Alsuraiheed, Elif Segah Oztas, Shakir Ali, Merve Bulut Yilgor . Reversible codes and applications to DNA codes over F42t[u]/(u21). AIMS Mathematics, 2023, 8(11): 27762-27774. doi: 10.3934/math.20231421
    [3] Ismail Aydogdu . On double cyclic codes over Z2+uZ2. AIMS Mathematics, 2024, 9(5): 11076-11091. doi: 10.3934/math.2024543
    [4] Ted Hurley . Ultimate linear block and convolutional codes. AIMS Mathematics, 2025, 10(4): 8398-8421. doi: 10.3934/math.2025387
    [5] Chaofeng Guan, Ruihu Li, Hao Song, Liangdong Lu, Husheng Li . Ternary quantum codes constructed from extremal self-dual codes and self-orthogonal codes. AIMS Mathematics, 2022, 7(4): 6516-6534. doi: 10.3934/math.2022363
    [6] Adel Alahmadi, Tamador Alihia, Patrick Solé . The build up construction for codes over a non-commutative non-unitary ring of order 9. AIMS Mathematics, 2024, 9(7): 18278-18307. doi: 10.3934/math.2024892
    [7] Hatoon Shoaib . Double circulant complementary dual codes over F4. AIMS Mathematics, 2023, 8(9): 21636-21643. doi: 10.3934/math.20231103
    [8] Adel Alahmadi, Altaf Alshuhail, Patrick Solé . The mass formula for self-orthogonal and self-dual codes over a non-unitary commutative ring. AIMS Mathematics, 2023, 8(10): 24367-24378. doi: 10.3934/math.20231242
    [9] Xuesong Si, Chuanze Niu . On skew cyclic codes over M2(F2). AIMS Mathematics, 2023, 8(10): 24434-24445. doi: 10.3934/math.20231246
    [10] Yuezhen Ren, Ruihu Li, Guanmin Guo . New entanglement-assisted quantum codes constructed from Hermitian LCD codes. AIMS Mathematics, 2023, 8(12): 30875-30881. doi: 10.3934/math.20231578
  • In obesity studies, several researchers have been applying machine learning tools to identify factors affecting human body weight. However, a proper review of strength, limitations and evaluation metrics of machine learning algorithms in obesity is lacking. This study reviews the status of application of machine learning algorithms in obesity studies and to identify strength and weaknesses of these methods. A scoping review of paper focusing on obesity was conducted. PubMed and Scopus databases were searched for the application of machine learning in obesity using different keywords. Only English papers in adult obesity between 2014 and 2019 were included. Also, only papers that focused on controllable factors (e.g., nutrition intake, dietary pattern and/or physical activity) were reviewed in depth. Papers on genetic or childhood obesity were excluded. Twenty reviewed papers used machine learning algorithms to identify the relationship between the contributing factors and obesity. Regression algorithms were widely applied. Other algorithms such as neural network, random forest and deep learning were less exploited. Limitations regarding data priori assumptions, overfitting and hyperparameter optimization were discussed. Performance metrics and validation techniques were identified. Machine learning applications are positively impacting obesity research. The nature and objective of a study and available data are key factors to consider in selecting the appropriate algorithms. The future research direction is to further explore and take advantage of the modern methods, i.e., neural network and deep learning, in obesity studies.



    Massey [1] introduced reversible codes that are known to exhibit useful properties in certain retrieval systems and data storage. Moreover, some reversible codes have been proven to possess excellent solid burst error correction capability and high transmission efficiency [2,3]. It has also recently been observed that reversible codes have broad applications in various areas of mathematics, including cryptography [4,5], and the theory of DNA codes [6,7,8].

    Interestingly, the class of reversible codes is closely connected to that of BCH codes because it is an important subclass of BCH codes. It is also closely related to the class of linear complementary dual codes (LCD codes, in short) introduced by Massey in [9]. In fact, Yang and Massey proved that a cyclic code is a reversible code if and only if it is an LCD code [10]. This interconnection between reversible codes, BCH codes, and LCD codes adds to their importance and potential applications in various domains.

    The Rosenbloom-Tsfasman metric, also known as the RT-metric, was first introduced by Rosenbloom and Tsfasman [11] in the field of coding theory. It was later introduced into the theory of uniform distributions by Skriganov [12] and Martin and Stinson [13]. RT-metric is a generalization of the classical Hamming metric, and it has immediately attracted the attention of many coding theorists, resulting in extensive research on codes equipped with this metric. The majority of research on codes in this metric focuses on various bounds [14], linearity [15,16,17], weight distribution and MacWilliam's identities [18,19,20,21], groups of automorphisms [22], maximum distance separability [23], burst error enumeration [24,25,26], normality [27], covering properties [28], construction of self-dual codes [29], and existence of LCD codes [30] over various algebraic structures.

    In the context of coding theory [11] and its corresponding notion in uniform distributions [12], the goal is to construct an RT-metric code with codewords that are maximally distant from each other, aiming for the smallest RT-distance between any two codewords to be as large as possible. Additionally, there is a desire for the RT-metric code to be as large as possible, akin to codes in the classical Hamming metric. However, these two objectives often conflict. Therefore, the aim is to achieve the maximum number of codewords with the greatest possible minimum distance or the largest minimum distance for a given number of codewords. Codes meeting these criteria are termed maximum distance separable (MDS) codes. Rosenbloom and Tsfasman [11] initially defined MDS codes over Fq with the ρ-metric in relation to potential information theoretic applications. Furthering their theory, Skriganov [12] related these codes to uniform distributions. An [n,k,dρ]q code in the RT-metric that attains the Singleton bound is considered an MDS code, meaning dρ=nk+1. Marka and Selvaraj [31] demonstrated that optimal codes in Fnq are MDS, and vice versa.

    As a result of this intriguing distinction of MDS codes in the RT metric, there arises a significant need for comprehensive study of MDS codes in this metric, such as the existence of MDS reversible codes. This paper aims to address this specific problem by investigating the presence of MDS reversible codes and subsequently exploring their properties if they are found to exist.

    Let u=(u1,u2,,un) and v=(v1,v2,,vn) be any two vectors in Fnq. The ρ-distance or RT-distance between them is denoted by dρ(u,v), defined as dρ(u,v)=max{i|uivi,1in}. A q-ary RT-metric code of length n refers to a subset of the space Fnq equipped with this metric. If this subset is a subspace, it is referred to as a linear RT-metric code. A generator matrix G of a k-dimensional linear code C in Fnq is a k×n matrix such that its rows form a basis for C. The coordinates of any set of k linearly independent columns of G represent the information set for code C.

    To derive MacWilliam's type relations for codes in the RT-metric, an essential inner product was introduced in [19] on the matrix space Matm×s(Fq). This particular inner product holds great importance in the investigation of codes in the RT-metric, as it influences various intriguing results. For instance, it indicates that the dual of an MDS code, under this specific inner product will also be an MDS code, which represents one of the noteworthy results in this context.

    For α=(α1,α2,...,αn),β=(β1,β2,...,βn)Fnq, the inner product of α and β is given by

    α,β=β,α=ni=1αiβni+1.

    Then, the dual C of the code C can be defined as

    C={αFnq|α,β=0forallβC}.

    An RT-metric code C is categorized based on certain properties. If C is contained within its dual code C, it is referred to as self-orthogonal. A self-dual code, on the other hand, satisfies the condition C=C. In contrast, for a code to be LCD, there are no non-zero codewords in common between C and its dual C. C is termed reversible if for every codeword c=(v1,v2,,vn) in C, its reverse Flip(c)=(vn,vn1,,v2,v1) is also an element of C and c is self-reversible if c=Flip(c).

    Let P=(pij)m×n be a matrix of size m×n. Then, we use the following notation (see Table 1) throughout this study.

    Table 1.  Notations and abbreviations.
    In the identity matrix of degree n
    Rn a matrix (rij)n×n, where rij=1, if i+j=n+1 and rij=0 otherwise
    PT the transpose of a matrix P, given by PT=(pji)n×m
    Flip(P) the column-reversed matrix of a matrix P, given by Flip(P)=(pi,nj+1)m×n
    RmP the row-reversed matrix of a rectangle matrix P
    PS the flip-diagonal transpose of a matrix P, which transposes the flip of P across its diagonal, i.e., PS=(pmi+1,nj+1)m×n
    Pk k×k square matrix
    RMDS reversible MDS
    SR self-dual reversible
    SRMDS self-dual reversible MDS
    x it rounds x up to the nearest integer

     | Show Table
    DownLoad: CSV

    Let P and Q be square matrices of order n. If PPT=In=PTP, then P is called orthogonal. If P=PT, then P is called symmetric. P is called centrosymmetric if P=PS. Furthermore, the following properties are straightforward:

    Rn=Flip(In);

    RTn=RSn=Rn;

    R2n=In;

    PS=RnPRn;

    (PT)S=(PS)T;

    (PS)S=P;

    (P+Q)S=PS+QS;

    (PQ)S=QSPS;

    Flip(P)=PRn.

    Lemma 3.1. If C is a reversible code of length n with dimension k over Fq with generator matrix G, then GS is also a generator matrix of C.

    Proof. Consider C as a reversible code with generator matrix G. Since C is reversible, Flip(G) is also a generator matrix of C. If Rk is non-singular, then GS = RkFlip(G) is also a generator matrix of C. Thus, the lemma holds true.

    Theorem 3.1. Let C be a linear code of even length n over Fq, with a generator matrix in the form G=[A|AS]. Then, C is reversible.

    Proof.

    G=[A|AS]isageneratormatrixofCFlip(G)=[Flip(AS)|Flip(A)]RkFlip(G)=Rk[Flip(AS)|Flip(A)]GS=[(AS)S|AS]GS=[A|AS]GS=G.

    Suppose C is not reversible. Then, Flip(G) cannot be a generator matrix of C. However, as Rk is non-singular, it implies that GS=RkFlip(G) is also not a generator matrix of C, leading to a contradiction. Therefore, C must be reversible.

    Theorem 3.2. Let C be a linear code of odd length n over Fq, with a generator matrix G in the form G=[A|y|AS], where y is a column vector such that y=yS. Then, C is reversible.

    Proof. The proof is similar to that of Theorem 3.1.

    Theorem 3.3. Let C be a reversible code of length n with dimension k over Fnq. Then:

    (i) If C has an odd length n, then we can express a generator matrix of C in the form G=[A|y|AS], where y is a column vector such that y=yS if and only if the total number of self-reversible codewords in code C is qk2.

    (ii) If C has an even length n, then we can represent a generator matrix of C in the form G=[A|AS] if and only if the total number of self-reversible codewords in C is qk2.

    Proof. Let C be a reversible code of length n with dimension k over Fnq.

    (ⅰ) Let us assume that G=[A|y|AS] represents a generator matrix for a reversible code C with an odd length n, where A is a matrix and y is a column vector satisfying y=yS.

    Case A: Assume k is even. Then, no self-reversible codeword exists as a row of generator matrix G=[A|y|AS]. Therefore, ri=Flip(rki+1),i{1,2,,k}, where ri is the ith row of G. Since C is reversible, ci=ri+rki+1(i=1,2,,k2) are k2 distinct self-reversible codewords. All distinct self-reversible codewords thus form a subspace of dimension k2 of C, leading to a total of qk2 self-reversible codewords.

    Case B: Assume k is odd. Then, exactly one self-reversible codeword exists as a row of generator matrix G=[A|y|AS] which is the (k+12)th row of G. Therefore, ri=Flip(rki+1),i{1,2,,k12}, where ri is the ith row of G. Since C is reversible, ci=ri+rki+1(i=1,2,,k12) and the (k+12)th row vectors as self-reversible codewords in G are k+12 distinct self-reversible codewords. Therefore, all distinct self-reversible codewords form a subspace of dimension k+12 of C, resulting in a total of qk+12 self-reversible codewords.

    From Cases A and B, we can conclude that the total number of self-reversible codewords in code C is qk2.

    Conversely, assume that the total number of self-reversible codewords in code C is qk2.

    Case A: Suppose k is even, and the total number of self-reversible codewords in C is qk2. Let U={u1,u2,,uk2} be a linearly independent subset of all self-reversible codewords in C. This subset forms a subspace with dimension k2 of C. Since C is reversible, every self-reversible codeword of subspace U can be written as ui=vi+Flip(vi)(i=1,2,,k2). Then, the sets Wi=<vi,Flip(vi)>, for i=1,2,,k2, span distinct subspaces of dimension 2 of C. For each j, WjijWi={0}, for i,j=1,2,,k2. Thus, B=k2i=1Wi, it contains k linearly independent codewords, and the direct sum of Wi's forms the basis of C, i.e., B=W1W2Wk2 is a basis of C.

    Case B: Suppose k is odd and the total number of self-reversible codewords in C is qk+12. Let t be one of the self-reversible codewords in C. Therefore, the subset t forms a subspace of dimension 1 of C. Let U={u1,u2,,uk12} be a linearly independent subset of all self-reversible codewords in C and it is disjoint from the subspace t in C. This subset forms a subspace with dimension k12 of C. Since C is reversible, every self-reversible codeword of subspace U can be written as ui=vi+Flip(vi)(i=1,2,,k12). Then, the sets Wi=<vi,Flip(vi)>,fori=1,2,,k12, span distinct subspaces of dimension 2 of C. For each j, WjijWi={0}, for i,j=1,2,,k12. Thus, B=k12i=1Wi, it contains k1 linearly independent codewords, and the direct sum of Wi's and t forms a basis of C, i.e., B=W1W2Wk12t is a basis of C.

    (ⅱ) The proof is similar to that for 3.3(i).

    Example 3.1. Consider a [7,3,4] reversible code C over GF(3) whose generator matrix is given by

    G=[121011200010002110121].

    The total number of self-reversible codewords in C is qk2=9 as the generators are (0,0,0,1,0,0,0) and (0,0,1,0,1,0,0). It is also to be observed that the generator matrix G above is in the form [A|y|AS]. This example follows from Theorem 3.3(i).

    Example 3.2. Consider a [6,4,2] reversible code C over GF(2) whose generator matrix is given by

    G=[110000101010010101000011].

    The total number of self-reversible codewords in C is qk2=4. It is also to be observed that the generator matrix G above is in the form G=[A|AS]. This example follows from Theorem 3.3(ii).

    Remark 3.1. Some of the reversible codes with a generator matrix G cannot be represented in the form G=[A|y|AS] or G=[A|AS], because the total number of self-reversible codewords in C is not equal to qk2. This can be seen from the following examples (Examples 3.3 and 3.4).

    Example 3.3. The matrix G given by

    G=[10000000001001000100001000000001]

    is a generator matrix for an [8,4,1] binary reversible code C in the RT-metric. However, this generator matrix G cannot be written in the form of [A|AS] for any matrix A. It is also to be noted that the total number of self-reversible codewords in C is 8, which is not equal to qk2=4.

    Example 3.4. The matrix G given by

    G=[101000102000101]

    is a generator matrix for a [5,3,3] ternary reversible code C in the RT-metric. However, this generator matrix G cannot be written in the form of G=[A|y|AS] for any matrix A. It is also to be noted that the total number of self-reversible codewords in C is 3, which is not equal to qk2=9.

    Theorem 3.4. Let C be a self-reversible code of length n with dimension k over Fnq. Then:

    (i) If C has an odd length n, then we can find a generator matrix of C in the form G=[A|y|Flip(A)], where y is a column vector, if and only if the total number of self-reversible codewords in C is qk.

    (ii) If C has an even length n, then we can find a generator matrix of C in the form G=[A|Flip(A)] if and only if the total number of self-reversible codewords in C is qk.

    Proof. The proof of this theorem is similar to the proof of Theorem 3.3.

    Theorem 3.5. Let C be an [n,k,d]q reversible code with the total number of self-reversible codewords qk2. Then, the dual C of C is an [n,nk,d]q reversible code, with the total number of self-reversible codewords qnk2.

    Proof. The proof of Theorem 3.5 is straightforward, relying on notations and basic algebraic manipulations.

    Theorem 4.1. Let C be a code of length n with dimension k (where kn2) in the RT-metric, whose generator matrix is in the form G=[Ak|Y|ASk], where Y is a k×(n2k) matrix. Then, C is MDS if and only if Ak is non-singular.

    Proof. Suppose that C is a code of length n with dimension k (where kn2) in the RT-metric, whose generator matrix is in the form G=[Ak|Y|ASk], where Y is a k×(n2k) matrix. Let C be an MDS code. Assume in contrary that Ak is singular. Then, ASk is also singular. Consequently, there exist at least two codewords x and z in C with the last k positions being the same, i.e., x=(x1,x2,,xnk,xnk+1,,xn) and z=(z1,z2,,znk,xnk+1,,xn). According to [31], "an (n,K,dρ)q code is MDS if and only if its partition number is ndρ+1." Thus, none of the two codewords have the same (ndρ+1)-tuple as their last ndρ+1 coordinates, which leads to a contradiction. Hence, Ak is non-singular.

    Conversely, assume that Ak is non-singular. Then, (ASk)1 is also non-singular. Thus, (ASk)1G is a generator matrix of C in the following form: (ASk)1G=[(ASk)1Ak|(ASk)1Y|(ASk)1ASk]=[(ASk)1Ak|(ASk)1Y|Ik]. Hence, C is an MDS.

    The following examples (Examples 5 and 6) illustrate Theorem 4.1.

    Example 4.1. Consider a [10,4,7] linear code C over GF(2) whose generator matrix is given by

    G=[1000100011001000101001010101001100110001].

    Here, d1=7, d2=8, d3=9, and d4=10. It is also to be observed that the generator matrix G above is in the form [Ak|Y|ASk] and that Ak is invertible.

    Example 4.2. Consider a [6,3,4] reversible code C over GF(3) whose generator matrix is given by

    G=[121122002200221121].

    Here, d1=4, d2=5, and d3=6. It is also to be observed that the generator matrix G above is in the form G=[Ak|ASk] and that Ak is invertible.

    Corollary 4.1. Let C be a code of length n with dimension k (where kn2) in the RT-metric, whose generator matrix is in the form G=[Ak|Y|ASk], where Y is a k×(n2k) matrix. Then, C is RMDS if and only if Ak is non-singular and Y is centrosymmetric.

    Proof. If we assume that C is RMDS in the RT-metric, from Theorem 4.1, Ak is non-singular. It is sufficient to prove that Y is centrosymmetric. As C is reversible, from Lemma 1, GS is also a generator matrix of C. Note that the rows of GS=[Ak|YS|ASk] and those of G=[Ak|Y|ASk] generate the same code C. This implies that Y=YS or Y is centrosymmetric.

    Conversely, assume that Ak is non-singular and Y is centrosymmetric. G=[Ak|Y|ASk] is a generator matrix of C, and (ASk)1G=[(ASk)1Ak|(ASk)1Y|(ASk)1ASk] is a generator matrix of C. This implies that G=[Bk|Y1|Ik] is a generator matrix of C, where Bk=(ASk)1Ak and Y1=(ASk)1Y. It is sufficient to prove that C is reversible, that is, to prove Y1=BkYS1 and (Flip(Bk))2=BkBSk=BSkBk=Ik. In [32], Theorem 3 states that the linear code C is RMDS code in the RT-metric if and only if Y1=BkYS1 and (Flip(Bk))2=BkBSk=BSkBk=Ik. Thus, Bk=(ASk)1AkandBSk=A1kR1kAkRk.

    ConsiderY1=(ASk)1YS(Yiscentrosymmetric)=(ASk)1(AkA1k)(RkFlip(Y))=((ASk)1Ak)(RkR1kA1k)(RkFlip(Y))=(Bk)Rk(RkAkRk)1(Flip(Y))(Bk=(ASk)1Ak)=(Bk)(Rk(ASk)1(Flip(Y)))=BkYS1(Y1=(ASk)1Y).
    Consider(Flip(Bk))2=BkBSk=((ASk)1Ak)(A1kR1kAkRk)=(RkAkRk)1(RkAkRk)=Ik.

    The following example (Example 4.3) illustrates Corollary 4.1.

    Example 4.3. Consider an [11,4,8] reversible code C over GF(2) whose generator matrix is given by

    G=[11100101000010001100111100110001000010100111].

    Here, d1=8, d2=9, d3=10, and d4=11. It is also to be observed that the generator matrix G above is in the form [Ak|Y|ASk]. Thus, Ak is invertible and Y is centrosymmetric.

    Theorem 4.2. Let C be a code of even length n with dimension k (n>k>n/2) in the RT-metric, where its generator matrix is in the form G=[A|AS], with Ak representing the first k×k square matrix of G. Then, the C is RMDS if and only if Ak is non-singular.

    Proof. The proof is similar to the proof of Theorem 4.1. The following example (Example 4.4) illustrates Theorem 4.2.

    Example 4.4. Consider a [6,4,3] reversible code C over GF(2) whose generator matrix is given by

    G=[100011001010010100110001].

    Here, d1=3, d2=4, d3=5, and d4=6. It is also to be observed that the generator matrix G above is in the form G=[A|AS] and that Ak is invertible.

    Theorem 4.3. Let C be a code of odd length n, with dimension k where n>k>n2, in the RT-metric. Its generator matrix is in the form G=[A|y|AS], where Ak represents the first k×k square matrix of G. Then, the C is RMDS if and only if Ak is non-singular and y is a column centrosymmetric vector.

    Proof. The proof is similar to the proof of Corollary 1. The following example (Example 4.5) illustrates Theorem 4.3.

    Example 4.5. Consider a [3,2,2] reversible code C over GF(5) whose generator matrix is given by

    G=[340043].

    Here, d1=2 and d2=3. It is also to be observed that the generator matrix G above is in the form G=[A|y|AS] and that Ak is invertible.

    Theorem 4.4. Every self-reversible [n=2k,k,dρ] code C in the RT-metric is MDS.

    Proof. Let C be any self-reversible [n=2k,k,dρ]q code with even length n=2k. Then, by using Definition 3 in [27], there exists a partition number l=k such that C is of type (k+1,k+2,2k). Hence, C is MDS. The following example (Example 4.6) illustrates Theorem 4.4.

    Example 4.6. Consider a [6,3,4] reversible code C over GF(2) whose generator matrix is given by

    G=[001100010010100001].

    Here, d1=4, d2=5, and d3=6. It is also to be observed that the generator matrix G above is in the form G=[Ak|Flip(Ak)] and that Ak is invertible.

    Theorem 4.5. Every self-reversible [n=2k1,k,dρ] code C with odd length n=2k1 in the RT-metric is MDS.

    Proof. The proof of this theorem is similar to the proof of Theorem 4.4. The following example (Example 4.7) illustrates Theorem 4.5.

    Example 4.7. Consider a [5,3,3] reversible code C over GF(3) whose generator matrix is given by

    G=[002000101020002].

    Here, d1=4, d2=5, and d3=6. It is also to be observed that the generator matrix G above is in the form G=[A|Flip(A)] and that Ak is invertible.

    Theorem 5.1. If C is an [n=2k,k,dρ] reversible code over Fq in the RT-metric with a minimum distance dρ=1, then C cannot be self-dual in the RT-metric.

    Proof. The proof of Theorem 5.1 is straightforward, relying on notations and basic algebraic manipulations.

    Remark 5.1. If C is an [n=2k,k,dρ] self-dual code over Fq in the RT-metric with a minimum distance of dρ=1, then C cannot be reversible. This can be seen from the following examples (Examples 5.1 and 5.2).

    Example 5.1. Consider a [6,3,1] linear code C over GF(2) whose generator matrix is given by

    G=[100000011000010110].

    Here, d1=1, d2=3, and d3=5. It can be observed that C is a [6,3,1] self-dual code in the RT-metric over GF(2), but it is not reversible.

    Example 5.2. Consider a [4,2,1] linear code C over GF(5) whose generator matrix is given by

    G=[40002030].

    Here, d1=1 and d2=3. It can be observed that C is a [4,2,1] self-dual code in the RT-metric over GF(5), but it is not reversible.

    Theorem 5.2. Let C be an [n=2k,k,dρ] binary reversible code in the RT-metric, with a generator matrix in the form [Ak|ASk]. Then, C is self-dual if and only if it satisfies one of the following conditions:

    (i) AkFlip(ATk)=(AkFlip(ATk))T;

    (ii) Flip(AkATk) is symmetric;

    (iii) AkATk is centrosymmetric.

    Proof. Consider:

    (i) Suppose C is self-dual [29], which implies that

    GG=0[Ak|ASk][(Flip(ASk))TFlip(Ak)T]=0Ak(Flip(ASk))T+ASk(Flip(Ak))T=0RkAkATk+AkFlip(ATk)=0RkAkATk=AkFlip(ATk)AkATkRk=(AkFlip(ATk))TAkFlip(ATk)=(AkFlip(ATk))T.

    (ii) Consider

    (AkATk)Rk=(AkFlip(ATk))TFlip(AkATk)=(AkATkRk)TFlip(AkATk)=(Flip(AkATk))TFlip(AkATk)is symmetric.

    (iii) Consider

    RkAkATk=AkFlip(ATk)RkAkATk=AkATkRkRk(AkATk)=(AkATk)Rk(AkATk)is centrosymmetric.

    Example 5.3. Consider a [6,3,3] reversible code C over GF(2) whose generator matrix is given by

    G=[101000010010000101].

    Here, d1=3, d2=5, and d3=6. It is also to be observed that the generator matrix G above is in the form G=[Ak|ASk] and that Flip(AkATk) is symmetric.

    Theorem 5.3. Let C be an [n=2k,k,dρ] non-binary reversible code in the RT-metric, with a generator matrix in the form [Ak|ASk]. Then, C is self-dual if and only if it satisfies one of the following conditions:

    (i) AkFlip(ATk)=(AkFlip(ATk))T;

    (ii) Flip(AkATk) is skew-symmetric.

    Proof. The proof of this theorem is similar to the proof of 5.2.

    Example 5.4. Consider a [4,2,3] reversible code C over GF(5) whose generator matrix is given by

    G=[40300304].

    Here, d1=3 and d2=4. It is also to be observed that the generator matrix G above is in the form G=[Ak|ASk] and that Flip(AkATk) is skew-symmetric.

    Theorem 5.4. Every binary self-reversible [n=2k,k,dρ] code C in the RT-metric is SR-MDS.

    Proof. The proof of Theorem 5.4 is straightforward, relying on notations and basic algebraic manipulations.

    Example 5.5. Consider a [4,2,3] binary self-reversible code C over GF(2) whose generator matrix is given by

    G=[10010110].

    Here, d1=3 and d2=4. It is also to be observed that the generator matrix G above is in the form G=[Ak|Flip(Ak)], and that C is self-dual.

    Theorem 6.1. Let G be a generator matrix of a self-dual code of length n. Then,

    [GOn/2,nOn/2,nFlip(G)]

    generates an SR code of length 2n in the RT-metric.

    Theorem 6.2. Let C be an [n1,k,d,R] RMDS code with a generator matrix in the form [AS|A] in the RT-metric. Then,

    [ORn2In2OASOOA]

    generates an RMDS code [2(n1+n2),n1+n2,dρ+n2,R+n2] with covering radius R+n.

    Proof. Consider

    [ORn2In2OASOOA][OFlip(A)TRn2OIn2OOFlip(AS)T]
    =[2In2OOAFlip(AT)+(AFlip(AT))T]
    =O(ifCisbinaryselfdual).

    In this study, we provided some basic properties of reversible linear codes and obtained a condition for a reversible code to be an MDS code. Furthermore, we established some necessary and sufficient conditions for a reversible code to be self-dual. Finally, some constructions of reversible codes in the RT-metric were proposed.

    Conceptualization: Bodigiri Sai Gopinadh formulated the initial research problem and developed the overarching mathematical framework. Validation: Venkatrajam Marka independently verified the correctness of the mathematical results, played a pivotal role in problem discussion, and providing continuous supervision. Both authors have read and approved the final version of the manuscript for publication.

    The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.

    The authors declare no conflicts of interest.



    [1] WHO, Obesity and Overweight, World Health Organization, 2020. Available from: https://wwwwhoint/news-room/fact-sheets/detail/obesity-and-overweight.
    [2] A. Hruby, J. E. Manson, L. Qi, V. S. Malik, E. B. Rimm, Q. Sun, W. C. Willett, F. B. Hu, Determinants and consequences of obesity, Am. J. Public Health, 106 (2016), 1656-1662. https://doi.org/https://doi.org/10.2105/AJPH.2016.303326 doi: 10.2105/AJPH.2016.303326
    [3] WHO, The top 10 causes of death, World Health Organization, 2018. Available from: https://wwwwhoint/news-room/fact-sheets/detail/the-top-10-causes-of-death.
    [4] WHO, 10 facts on obesity, World Health Organization, 2017. Available from: https://wwwwhoint/features/factfiles/obesity/en/..
    [5] J. Cawley, C. Meyerhoefer, The medical care costs of obesity: An instrumental variables approach, J. Health Econ., 31 (2012), 219-230. https://doi.org/10.1016/j.jhealeco.2011.10.003 doi: 10.1016/j.jhealeco.2011.10.003
    [6] L. Angrisani, A. Santonicola, P. Iovino, G. Formisani, H. Buchwald, N. Scopinaro, Bariatric Surgery Worldwide 2013, Obes. Surg., 25 (2015), 1822-1832. https://doi.org/10.1007/s11695-015-1657-z doi: 10.1007/s11695-015-1657-z
    [7] T. Bhurosy, R. Jeewon, Overweight and obesity epidemic in developing countries: A problem with diet, physical activity, or socioeconomic status? Scientific World Journal, 2014 (2014). https://doi.org/10.1155/2014/964236
    [8] E. Alpaydin, Introduction to Machine Learning, Cambridge: MIT press, 2014.
    [9] N. S. Rajliwall, R. Davey, G. Chetty, Machine learning based models for cardiovascular risk prediction, International Conference on Machine Learning and Data Engineering 2018, (iCMLDE), (2018), 142-148. https://doi.org/10.1109/iCMLDE.2018.00034
    [10] J. B. Heaton, N. G. Polson, J. H. Witte, Deep learning for finance: deep portfolios, Appl. Stoch. Model. Bus., 33 (2017), 3-12. https://doi.org/10.1002/asmb.2209 doi: 10.1002/asmb.2209
    [11] J. Kim, J. Canny, Interpretable Learning for Self-Driving Cars by Visualizing Causal Attention, Proceedings of the IEEE International Conference on Computer Vision, (2017), 2942-2950. https://doi.org/10.1109/ICCV.2017.320
    [12] D. Gruson, T. Helleputte, P. Rousseau, D. Gruson, Data science, artificial intelligence, and machine learning: Opportunities for laboratory medicine and the value of positive regulation, Clin. Biochem., 69 (2019), 1-7. https://doi.org/10.1016/j.clinbiochem.2019.04.013 doi: 10.1016/j.clinbiochem.2019.04.013
    [13] D. Panaretos, E. Koloverou, A. C. Dimopoulos, G. M. Kouli, M. Vamvakari, G. Tzavelas, C. Pitsavos, D. B. Panagiotakos, A comparison of statistical and machine-learning techniques in evaluating the association between dietary patterns and 10-year cardiometabolic risk (2002-2012): The ATTICA study, Brit. J. Nutr., 120 (2018), 326-334. https://doi.org/10.1017/S0007114518001150 doi: 10.1017/S0007114518001150
    [14] H. C. Koh, G. Tan, Data Mining Applications in Healthcare, Journal of Healthcare Information Management, 19 (2011), 64-72.
    [15] K. Kourou, T. P. Exarchos, K. P. Exarchos, M. V. Karamouzis, D. I. Fotiadis, Machine learning applications in cancer prognosis and prediction, Comput. Struct. Biotec., 13 (2015), 8-17. https://doi.org/10.1016/j.csbj.2014.11.005 doi: 10.1016/j.csbj.2014.11.005
    [16] V. Gulshan, L. Peng, M. Coram, M. C. Stumpe, D. Wu, A. Narayanaswamy, S. Venugopalan, K. Widner, et al., Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs, JAMA - Journal of the American Medical Association, 316 (2016), 2402-2410. https://doi.org/10.1001/jama.2016.17216 doi: 10.1001/jama.2016.17216
    [17] Y. Xing, J. Wang, Z. Zhao, Combination data mining methods with new medical data to predicting outcome of Coronary Heart Disease, International Conference on Convergence Information Technology, (ICCIT) 2007, (2007), 868-872. https://doi.org/10.1109/ICCIT.2007.4420369
    [18] P. Fränti, S. Sieranoja, K. Wikströ m, T. Laatikainen, Clustering diagnoses from 58M patient visits in Finland during 2015-2018, JMIR Medical Informatics, (2022). https://doi.org/10.2196/35422
    [19] Z. Obermeyer, E. J. Emanuel, Predicting the Future: Big Data, Machine Learning, and Clinical Medicine, The New England journal of medicine, 375 (2016), 1216-1219. https://doi.org/doi:10.1056/NEJMp1606181 doi: 10.1056/NEJMp1606181
    [20] M. A. Morris, E. Wilkins, K. A. Timmins, M. Bryant, M. Birkin, C. Griffiths, Can big data solve a big problem? Reporting the obesity data landscape in line with the Foresight obesity system map, Int. J. Obesity, 42 (2018), 1963-1976. https://doi.org/10.1038/s41366-018-0184-0 doi: 10.1038/s41366-018-0184-0
    [21] C. Y. J. Peng, K. L. Lee, G. M. Ingersoll, An introduction to logistic regression analysis and reporting, J. Educ. Res., 96 (2002), 3-14. https://doi.org/10.1080/00220670209598786 doi: 10.1080/00220670209598786
    [22] D. Dietrich, B. Heller, Y. Beibei, Data Science and Big Data Analytics: Discovering, Analyzing, Visualizing and Presenting Data, Indianapolis: Wiley, 2015.
    [23] H. O. Alanazi, A. H. Abdullah, K. N. Qureshi, A Critical Review for Developing Accurate and Dynamic Predictive Models Using Machine Learning Methods in Medicine and Health Care, J. Med. Syst., 41 (2017), 1-10. https://doi.org/10.1007/s10916-017-0715-6 doi: 10.1007/s10916-017-0715-6
    [24] Y. Y. Song, L. U. Ying, Decision tree methods: applications for classification and prediction, Shanghai Archives of Psychiatry, 27 (2015), 130-135. https://doi.org/10.11919/j.issn.1002-0829.215044 doi: 10.11919/j.issn.1002-0829.215044
    [25] M. Pal, Random forest classifier for remote sensing classification, Int. J. Remote Sens., 26 (2005), 217-222. https://doi.org/10.1080/01431160412331269698 doi: 10.1080/01431160412331269698
    [26] S. V. Vishwanathan, M. N. Murty, SSVM: A simple SVM algorithm, International Joint Conference on Neural Networks (IJCNN) 2002, 3 (2002), 2393-2398. https://doi.org/10.1109/IJCNN.2002.1007516
    [27] Y. Qu, B. Fang, W. Zhang, R. Tang, M. Niu, H. Guo, Y. Yu, X. He, Product-Based Neural Networks for User Response Prediction over Multi-Field Categorical Data, ACM T. Inform. Syst., 37 (2019), 1-35. https://doi.org/10.1145/3233770 doi: 10.1145/3233770
    [28] T. Chen, C. Guestrin, XGBoost: A scalable tree boosting system, Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, (2016), 785-794. https://doi.org/10.1145/2939672.2939785
    [29] A. T. C. Goh, Back-propagation neural networks for modeling complex systems, Artificial Intelligence in Engineering, 9 (1995), 143-151. https://doi.org/10.1016/0954-1810(94)00011-S doi: 10.1016/0954-1810(94)00011-S
    [30] Y. Lecun, Y. Bengio, G. Hinton, Deep learning, Nature, 521 (2015), 436-444. https://doi.org/10.1038/nature14539 doi: 10.1038/nature14539
    [31] A. K. Jain, M. N. Murty, P. J. Flynn, Data clustering: A review, ACM Comput. Surv., 31 (1999), 264-323. https://doi.org/10.1145/331499.331504 doi: 10.1145/331499.331504
    [32] H. Arksey, L. O'Malley, Scoping studies: towards a methodological framework, Int. J. Soc. Res. Method., 8 (2005), 19-32. https://doi.org/10.1080/1364557032000119616 doi: 10.1080/1364557032000119616
    [33] H. So, L. McLaren, G. C. Currie, The relationship between health eating and overweight/obesity in Canada: cross-sectional study using the CCHS, Obesity Science and Practice, 3 (2017), 399-406. https://doi.org/10.1002/osp4.123 doi: 10.1002/osp4.123
    [34] N. Daud, N. L. Mohd Noor, S. A. Aljunid, N. Noordin, N. I. M. F. Teng, Predictive Analytics: The Application of J48 Algorithm on Grocery Data to Predict Obesity, 2018 IEEE Conference on Big Data and Analytics, ICBDA, (2018), 1-6. https://doi.org/10.1109/ICBDAA.2018.8629623
    [35] J. F. Easton, H. Román Sicilia, C. R. Stephens, Classification of diagnostic subcategories for obesity and diabetes based on eating patterns, Nutr. Diet., 76 (2019), 104-109. https://doi.org/10.1111/1747-0080.12495 doi: 10.1111/1747-0080.12495
    [36] J. Dunstan, M. Aguirre, M. Bastías, C. Nau, T. A. Glass, F. Tobar, Predicting nationwide obesity from food sales using machine learning, Health Inform. J., 26 (2019), 652-663. https://doi.org/10.1177/1460458219845959 doi: 10.1177/1460458219845959
    [37] N. Kanerva, J. Kontto, M. Erkkola, J. Nevalainen, S. Mannisto, Suitability of random forest analysis for epidemiological research: Exploring sociodemographic and lifestyle-related risk factors of overweight in a cross-sectional design, Scand. J. Public Health, 46 (2018), 557-564. https://doi.org/10.1177/1403494817736944 doi: 10.1177/1403494817736944
    [38] K. W. DeGregory, P. Kuiper, T. DeSilvio, J. D. Pleuss, R. Miller, J. W. Roginski, C. B. Fisher, D. Harness, et al., A review of machine learning in obesity, Obes. Rev., 19 (2018), 668-685. https://doi.org/10.1111/obr.12667 doi: 10.1111/obr.12667
    [39] D. Kim, W. Hou, F. Wang, C. Arcan, Factors Affecting Obesity and Waist Circumference Among US Adults, Prev. Chronic Dis., 16 (2019). https://doi.org/10.5888/pcd16.180220
    [40] R. L. Figueroa, C. A. Flores, Extracting Information from Electronic Medical Records to Identify the Obesity Status of a Patient Based on Comorbidities and Bodyweight Measures, J. Med. Syst., 40 (2016). https://doi.org/10.1007/s10916-016-0548-8 doi: 10.1007/s10916-016-0548-8
    [41] M. A. Green, M. Strong, F. Razak, S. V. Subramanian, C. Relton, P. Bissell, Who are the obese? A cluster analysis exploring subgroups of the obese, J. Public Health (UK), 38 (2016), 258-264. https://doi.org/10.1093/pubmed/fdv040 doi: 10.1093/pubmed/fdv040
    [42] P. P. Brzan, Z. Obradovic, G. Stiglic, Contribution of temporal data to predictive performance in 30-day readmission of morbidly obese patients, PeerJ, 5 (2017), e3230. https://doi.org/10.7717/peerj.3230 doi: 10.7717/peerj.3230
    [43] A. Kupusinac, E. Stokić, R. Doroslovački, Predicting body fat percentage based on gender, age and BMI by using artificial neural networks, Comput. Meth. Prog. Bio., 113 (2014), 610-619. https://doi.org/10.1016/j.cmpb.2013.10.013 doi: 10.1016/j.cmpb.2013.10.013
    [44] M. Batterham, L. Tapsell, K. Charlton, J. O'shea, R. Thorne, Using data mining to predict success in a weight loss trial, J. Hum. Nutr. Diet., 30 (2017), 471-478. https://doi.org/10.1111/jhn.12448 doi: 10.1111/jhn.12448
    [45] Z. Feng, L. Mo, M. Li, A Random Forest-based ensemble method for activity recognition, 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2015 EMBS, (2015), 5074-5077. https://doi.org/10.1109/EMBC.2015.7319532
    [46] M. Batterham, E. Neale, A. Martin, L. Tapsell, Data mining: Potential applications in research on nutrition and health, Nutr. Diet., 74 (2017), 3-10. https://doi.org/10.1111/1747-0080.12337 doi: 10.1111/1747-0080.12337
    [47] W. J. Heerman, N. Jackson, M. Hargreaves, S. A. Mulvaney, D. Schlundt, K. A. Wallston, R. L. Rothman, Clusters of Healthy and Unhealthy Eating Behaviors Are Associated With Body Mass Index Among Adults, J. Nutr. Educ. Behav., 49 (2017), 415-421. https://doi.org/10.1016/j.jneb.2017.02.001 doi: 10.1016/j.jneb.2017.02.001
    [48] I. Sarasfis, C. Diou, I. Ioakimidis, A. Delopoulos, Assessment of In-Meal Eating Behaviour using Fuzzy SVM, 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), (2019), 6939-6942. https://doi.org/10.1109/EMBC.2019.8857606
    [49] P. Pouladzadeh, S. Shirmohammadi, A. Bakirov, A. Bulut, A. Yassine, Cloud-based SVM for food categorization, Multimed. Tools Appl., 74 (2015), 5243-5260. https://doi.org/10.1007/s11042-014-2116-x doi: 10.1007/s11042-014-2116-x
    [50] E. J. Heravi, H. Habibi Aghdam, D. Puig, A deep convolutional neural network for recognizing foods, Eighth International Conference on Machine Vision (ICMV), 9875 (2015), 98751D. https://doi.org/10.1117/12.2228875 doi: 10.1117/12.2228875
    [51] E. Disse, S. Ledoux, C. Bétry, C. Caussy, C. Maitrepierre, M. Coupaye, M. Laville, C. Simon, An artificial neural network to predict resting energy expenditure in obesity, Clin. Nutr., 37 (2018), 1661-1669. https://doi.org/10.1016/j.clnu.2017.07.017 doi: 10.1016/j.clnu.2017.07.017
    [52] N. Cesare, P. Dwivedi, Q. C. Nguyen, E. O. Nsoesie, Use of social media, search queries, and demographic data to assess obesity prevalence in the United States, Palgrave Communications, 5 (2019), 1-9. https://doi.org/10.1057/s41599-019-0314-x doi: 10.1057/s41599-019-0314-x
    [53] P. Kuhad, A. Yassine, S. Shimohammadi, Using distance estimation and deep learning to simplify calibration in food calorie measurement, IEEE International Conference on Computational Intelligence and Virtual Environments for Measurement Systems and Applications, CIVEMSA, (2015), 1-6. https://doi.org/10.1109/CIVEMSA.2015.7158594
    [54] K. Shameer, K. W. Johnson, B. S. Glicksberg, J. T. Dudley, P. P. Sengupta, Machine learning in cardiovascular medicine: Are we there yet? Heart, 104 (2018), 1156-1164. https://doi.org/10.1136/heartjnl-2017-311198 doi: 10.1136/heartjnl-2017-311198
    [55] B. A. Goldstein, A. M. Navar, R. E. Carter, Moving beyond regression techniques in cardiovascular risk prediction: Applying machine learning to address analytic challenges, Eur. Heart J., 38 (2017), 1805-1814. https://doi.org/10.1093/eurheartj/ehw302 doi: 10.1093/eurheartj/ehw302
    [56] N. Jothi, N. A. A. Rashid, W. Husain, Data Mining in Healthcare - A Review, Procedia Computer Science, 72 (2015), 306-313. https://doi.org/10.1016/j.procs.2015.12.145 doi: 10.1016/j.procs.2015.12.145
    [57] A. L. Beam, I. S. Kohane, Big data and machine learning in health care, JAMA - Journal of the American Medical Association, 319 (2018), 1317-1318. https://doi.org/10.1001/jama.2017.18391 doi: 10.1001/jama.2017.18391
    [58] A. Mozumdar, G. Liguori, Corrective Equations to Self-Reported Height and Weight for Obesity Estimates among U.S. Adults: NHANES 1999-2008, Res. Q. Exercise Sport, 87 (2016), 47-58. https://doi.org/10.1080/02701367.2015.1124971 doi: 10.1080/02701367.2015.1124971
    [59] M. Stommel, C. A. Schoenborn, Accuracy and usefulness of BMI measures based on self-reported weight and height: Findings from the NHANES & NHIS 2001-2006, BMC Public Health, 9 (2009), 1-10. https://doi.org/10.1186/1471-2458-9-421 doi: 10.1186/1471-2458-9-421
    [60] D. Rativa, B. J. T. Fernandes, A. Roque, Height and Weight Estimation from Anthropometric Measurements Using Machine Learning Regressions, IEEE J. Transl. Eng. He., 6 (2018), 1-9. https://doi.org/10.1109/JTEHM.2018.2797983 doi: 10.1109/JTEHM.2018.2797983
    [61] J. A. Sáez, J. Luengo, F. Herrera, Predicting noise filtering efficacy with data complexity measures for nearest neighbor classification, Pattern Recogn., 46 (2013), 355-364. https://doi.org/10.1016/j.patcog.2012.07.009 doi: 10.1016/j.patcog.2012.07.009
    [62] T. Ferenci, L. Kovács, Predicting body fat percentage from anthropometric and laboratory measurements using artificial neural networks, Applied Soft Computing Journal, 67 (2018), 834-839. https://doi.org/10.1016/j.asoc.2017.05.063 doi: 10.1016/j.asoc.2017.05.063
    [63] S. P. Goldstein, F. Zhang, J. G. Thomas, M. L. Butryn, J. D. Herbert, E. M. Forman, Application of Machine Learning to Predict Dietary Lapses During Weight Loss, Journal of Diabetes Science and Technology, 12 (2018), 1045-1052. https://doi.org/10.1177/1932296818775757 doi: 10.1177/1932296818775757
  • This article has been cited by:

    1. Bodigiri Sai Gopinadh, Venkatrajam Marka, Reversible codes in the Rosenbloom-Tsfasman metric, 2024, 9, 2473-6988, 22927, 10.3934/math.20241115
  • Reader Comments
  • © 2022 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(4362) PDF downloads(261) Cited by(11)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog