PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of springeropenLink to Publisher's site
Journal of Inequalities and Applications
 
J Inequal Appl. 2017; 2017(1): 300.
Published online 2017 December 6. doi:  10.1186/s13660-017-1541-6
PMCID: PMC5719149

A novel sequence space related to p defined by Orlicz function with application in pattern recognition

Abstract

In the field of pattern recognition, clustering groups the data into different clusters on the basis of similarity among them. Many a time, the similarity level between data points is derived through a distance measure; so, a number of clustering techniques reliant on such a measure are developed. Clustering algorithms are modified by employing an appropriate distance measure due to the high versatility of a data set. The distance measure becomes appropriate in clustering algorithm if weights assigned at the components of the distance measure are in concurrence to the problem. In this paper, we propose a new sequence space ℳ(ϕp, ℱ) related to p using an Orlicz function. Many interesting properties of the sequence space ℳ(ϕp, ℱ) are established by the help of a distance measure, which is also used to modify the k-means clustering algorithm. To show the efficacy of the modified k-means clustering algorithm over the standard k-means clustering algorithm, we have implemented them for two real-world data set, viz. a two-moon data set and a path-based data set (borrowed from the UCI repository). The clustering accuracy obtained by our proposed clustering algoritm outperformes the standard k-means clustering algorithm.

Keywords: clustering, double sequence, k-means clustering, Orlicz function

Introduction

Clustering is the process of separating a data set into different groups (clusters) such that objects in the same cluster should be similar to one another but dissimilar in another cluster [13]. It is a procedure to handle unsupervised learning problems appearing in pattern recognition. The major contribution in the field of clustering came due to the pioneering work of MacQueen [1] and Bazdek [2]. The k-means clustering algorithm was introduced by MacQueen [1], which is based on the minimum distance of the points from the center. The variants of k-means clustering algorithms were proposed to solve different types of pattern recognitions problems (see [47]). The clustering results of k-means or its variant can be further enhanced by choosing an appropriate distance measure. Therefore, the distance measure has a vital role in the clustering.

Clustering process is usually carried out through the l2 distance measure [8], but, due to its trajectory, sometimes it fails to offer good results. Suppose that two points x and y are selected on the boundary of the square (case p = 1) and let z be the center (Figure 1). Then l1 will fail to distinguish x and y, but these points may be distinguished by l2. If the points x and y are on the circumference of the circle, then l2 will fail to distinguish them. Moreover, the lp (p ≥ 1) distance measures are not flexible, so they cannot be modified as per the need of the clustering problem. Hence, clustering results derived through distance-dependent algorithms basically depend upon two properties of a distance measure: (1) trajectory and (2) flexibility. Till now, we have not come across to any distance measure that offers a guaranteed good result for every clustering problems. Clustering is carried out by using other variants of the lp distance measure. The distance measure of the sequence space lp,q, 1 ≤ pq ≤ ∞, introduced by Kellogg [9] and further studied by Jovanovic and Rakocevic [10], Oscar and Carme [11], and Ivana et al. [12] offers more flexibility in comparison to lp due to involvement of additional parameter q. Sargent [13] introduced another interesting sequence spaces m(φ) and n(φ) closely related to lp. Some useful extensions of m(φ) and n(φ) sequence spaces were proposed by Tripathy and Sen [14], Mursaleen [15, 16], and Vakeel [17]. Malkowsky et al. [18] defined a matrix mapping into the strong Cesàro sequence space [19] and studied the modulus function. Recently, for first time, Khan et al. [20] defined a distance measure of the double sequence of ℳ(ϕ) and 𝒩(ϕ) to cluster the objects. Moreover, Khan et al. in [38, 39] defined some more similarity measures by using distance measures of the double sequences in the uncertain environment. Mohiuddine and Alotaibi applied measures of noncompactness to solve an infinite system of second-order differential equations in p spaces [21, 22]. The double sequence space is further studied by Mursaleen and Mohiuddine [23], Altay and Başar [24, 25], Başar and Şever [26], and Esi and Hazarika [27]. Moreover, an Orlicz function and a fuzzy set are also used to define other types of double sequence spaces [18, 2831]. The convergence of difference sequence spaces is discussed in [29, 32, 33].

Figure 1
Geometry of lp norm.

In this paper, we define a new double sequence space ℳ(ϕp, ℱ) related to p using the following Orlicz function:

M(ϕ,p,F)={x={xmn}Ω:sups,t1supζUst1ϕstm,nζ(F(|xmn|ρ))p<, for some ρ>0}.

Obviously, ℳ(ϕp, ℱ) is a norm space, and hence the induced distance measure is represented as

dM(ϕ,p,F)(x,y)=(sups,t1supζUst1ϕstm,nζ(F(|xmnymn|ρ))p)1p.

The parameters ϕ, p and Orlicz function of ℳ(ϕp, ℱ) brings flexibility in the induced distance measure dM(ϕ,p,ℱ), which helps the user to modify it as per need of the clustering problem. Besides, defining the distance measure of ℳ(ϕp, ℱ), we have also studied some of its mathematically established properties. Finally, the distance measure of ℳ(ϕp, ℱ) is used in the k-means clustering algorithm, which clusters real-world data sets such as two-moon data set and path-based data set. The clustering results obtained by the modified clustering algorithm is compared with the k-means clustering algorithm to show its efficacy.

Preliminaries

Throughout the paper, l, c, and c0 denote the Banach spaces of bounded, convergent, and null sequences; ω, , and denote the sets of real (ordinary or single) sequences, natural numbers, and real numbers, respectively.

Orlicz function [34]

A function ℱ:[0, ∞) → [0, ∞) is called an Orlicz function if

  • (i) ℱ(0) = 0, ℱ(x) > 0 for x > 0, and ℱ(x) → ∞ as x → ∞;
  • (ii) is convex;
  • (iii) is nondecreasing; and
  • (iv) is continuous from the right of 0.

An Orlicz function is said to satisfy Δ2-condition for all values of x if there exists a constant K > 0 such that ℱ(2x) ≤ Kℱ(x) for all x ≥ 0. The Δ2-condition is equivalent to ℱ(Lx) ≤ Kℱ(x) for all values of x > 0 and for L > 1. An Orlicz function can always be represented in the following integral form:

F(x)=0xη(t)dt,

where η, known as the kernel of , is right-differentiable for t ≥ 0, η(0) = 0, η(t) > 0 for t > 0, η is nondecreasing, and η(t) → ∞ as t → ∞.

Let 𝒞 be the space of finite sets of distinct positive integers. Given any element σ of 𝒞. Let c(σ) be the sequence {cn(σ)} such that cn(σ) = 1 if n ∈ σ and cn(σ) = 0 otherwise. Further, let

Cs={σC:n=1cn(σ)s}(cf. [6])

be the set of those σ whose support has cardinality at most s, and

Φ={ϕ={ϕn}ω:ϕ1>0,Δϕn0 and Δ(ϕnn)0(n=1,2,)},

where Δφnφn − φn−1.

For φ ∈ Φ, the sequence space, introduced by Sargent [13] and known as Sargent’s sequence space, is defined as follows:

m(ϕ)={x={xn}ω:sups1supσCs(1ϕsnσ|xn|)<}.

Let Ω be the set of all real-valued double sequences, which is a vector space with coordinatewise addition and scalar multiplication. A double sequence x = {xmn} of real numbers is said to be bounded if x = supm,n|xmn| < ∞. We denote the space of all bounded double sequences by . Consider a sequence x = {xmn} ∈ Ω. If for every ε > 0, there exist nn(ε) ∈ ℕ and  ∈ ℝ such that

|xmn − | < ε

for all mn > n then we say that the double sequence x is convergent in the Pringheim sense to the limit [ell] and write 𝒫-lim xmn. By 𝒞p we denote the space of all convergent double sequences in the Pringsheim sense. It is well known that there are such sequences in the space 𝒞p but not in the space . So, we can consider the space 𝒞bp of double sequences that are both convergent in the Pringsheim sense and bounded, that is, 𝒞bp = 𝒞p ∩ ℒ. A double sequence x = {xmn} is said to converge regularly to [ell] (shortly, r-convergent to [ell]) if x is 𝒫-convergent to [ell] and the limits xm: = limnxm,n (m ∈ ℕ) and xn: = limmxm,n (n ∈ ℕ) exist. Note that, in this case, the limits limmlimnxm,n and limnlimmxm,n exist and are equal to the 𝒫-limit of x. Therefore, [ell] is called the r-limit of x.

In general, for any notion of convergence ν, the space of all ν-convergent double sequences will be denoted by 𝒞ν, and the limit of a ν-convergent double sequence x by ν-limm,nxmn, where ν ∈ {𝒫, bpr}.

Başar and Sever [26] have introduced the space p of p-summable double sequences corresponding to the space lp (p ≥ 1) of single sequences as

Lp:=({xmn}Ω:m,n|xmn|p<)(1p<)

and examined some properties of the space. Altay and Başar [25] have generalized the set of double sequences , 𝒞p, and 𝒞bp etc. by defining (t) = ({xmn} ∈ Ω:supm,n∈ℕ|xmn|tmn < ∞), 𝒞p(t) = ({xmn} ∈ Ω:𝒫-limm,n→∞|xmn − |tmn < ∞), and 𝒞bp(t) = 𝒞p ∩ ℒ, respectively, where t = {tmn} is a sequence of strictly positive reals tmn. In the case tmn = 1 for all mn ∈ ℕ, (t), 𝒞p(t), and 𝒞bp(t) reduce to the sets , 𝒞p and 𝒞bp, respectively.

Now just to have a better idea about other convergences, especially the linear convergence, we first consider the isomorphism defined by Zelster [35] as

T:Ωω,xz=(zi):=(xχ1(i)),
1

where χ:ℕ × ℕ → ℕ is the bijection defined by

χ[(1,1)]=1,χ[(1,2)]=2,χ[(2,2)]=3,χ[(2,1)]=4χ[(1,n)]=(n1)2+1,χ[(2,n)]=(n1)2+2,,χ[(n,n)]=(n1)2+n,χ[(n,n1)]=n2n+2,,χ[(n,1)]=n2,

Let us consider a double sequence x = {xmn} and define the sequence s = {smn} via x by

smn:=i,jm,nxij(m,nN).

For brevity, here and in what follows, we abbreviate the summations k=1l=1 and k=1ml=1n by i,j=1, and i,j=1m,n, respectively. Then the pair (xs) and the sequence s = {smn} are called a double series and the sequence of partial sums of a double series, respectively. Let λ be the space of double sequences, converging with respect to some linear convergence rule μ-lim :λ → ℝ. The sum of a double series i,j=1,xij with respect to this rule is defined by μ-i,j=1,xij:=μ-limsmn.

In this paper, we define an analogoue of Sargent’s sequence in the double sequence space Ω. For this, we first suppose that 𝒰 is the space whose elements are finite sets of distinct elements of ℕ × ℕ obtained by σ × ς, where σ ∈ 𝒞s and ς ∈ 𝒞t for each st ≥ 1. Therefore any element ζ of 𝒰 means (jk); j ∈ σ & k ∈ ς having cardinality atmost st, where s is the cardinality with respect to m, and t is the cardinality with respect to n. Here, the product say c of st may be same for differnt sets of positive integers kl, but in that case, 𝒰kl is different from 𝒰st. Given any element ζ of 𝒰, we denote by c(ζ) the sequence {cmn(ζ)} such that

cmn(ζ)={1if (m,n)ζ,0otherwise.

Further, let

Ust={ζU:m,n=1,cmn(ζ)st}

be the set of those ζ whose support has cardinality at most st, and let

Θ={ϕ={ϕmn}Ω:ϕ11>0,Δ10ϕmn,Δ01ϕmn,Δ11ϕmn0 andΔ10(ϕmnmn),Δ01(ϕmnmn),Δ11(ϕmnmn)0(m,n=1,2,)},

where Δ10φmnφmn − φm−1n, Δ01φmnφmn − φmn−1, Δ11φmnφmn − φm−1n−1.

For φ ∈ Θ, we define the sequence space

M(ϕ,F)={x={xmn}Ω:sups,t1supζUst1ϕstm,nζF(|xmn|ρ)<, for some ρ>0}.

Throughout the paper, m,nζ means mσnς.

The spaces ℳ(ϕ, ℱ), p, and can be extended to ℳ(ϕp, ℱ), p(ℱ), and (ℱ) as follows:

M(ϕ,p,F)={x={xmn}Ω:sups,t1supζUst1ϕstm,nζ(F(|xmn|ρ))p<, for some ρ>0},Lp(F)={{xmn}Ω:m,n(F(|xmn|ρ))p<, for some ρ>0}(1p<),L(F)={{xmn}Ω:supm,nF(|xmn|ρ)<, for some ρ>0}.

Now, if we take the cardinality t with respect to n as 1, then ℳ(ϕ, ℱ) reduce to m(ϕ, ℱ), and ℳ(ϕp, ℱ) to m(ϕp, ℱ). Here, without further discussing ℳ(φF), we immediately define ℳ(ϕp, ℱ) so as not to deviate from our main goal to show that ℳ(ϕp, ℱ) is a class of new double sequences lying between p(ℱ) and (ℱ). We then further prove certain conditions under which ℳ(ϕp, ℱ) is same as that of p(ℱ) and (ℱ). We can easily see that all results in Section 2 hold for ℳ(ϕ, ℱ), which is a particular case of ℳ(ϕp, ℱ) with p = 1.

Some interesting results related to ℳ(ϕp, ℱ)

Theorem 3.1

The sequence space ℳ(ϕp, ℱ) is a linear space over .

Proof

Let xy ∈ ℳ(ϕp, ℱ) and λμ ∈ ℝ. Then there exists positive numbers ρ1 and ρ2 such that

sups,t1supζUst1ϕstm,nζ(F(|xmn|ρ1))p<

and

sups,t1supζUst1ϕstm,nζ(F(|xmn|ρ2))p<.

Let ρ3 = max (2|λ|ρ1, 2|μ|ρ2).

(1) 0 < p < 1. Using the well-known inequality |ab|p ≤ |a|p + |b|p for 0 < p < 1 and the convexity of Orlicz functions, we have

sups,t1supζUst1ϕstm,nζ(F(|λxmn+μymn|ρ3))psups,t1supζUst1ϕstm,nζ(F(|λxmn|ρ3))p+sups,t1supζUst1ϕstm,nζ(F(|μymn|ρ3))psups,t1supζUst1ϕstm,nζ(F(|λxmn|ρ1))p+sups,t1supζUst1Ustm,nζ(F(|μymn|ρ2))p<,

so that λxmnμymn ∈ ℳ(ϕp, ℱ). This proves that ℳ(ϕp, ℱ) is a linear space over and so obviously is nonempty.

(2) 1 ≤ p <  + ∞. It is easy to see that for all ab ∈ ℝ, |ab|p ≤ 2p(|a|p + |b|p) and is convex, so that, for all st ≥ 1, ζ ∈ 𝒰st,

sups,t1supζUst1ϕstm,nζ(F(|λxmn+μymn|ρ3))psups,t1supζUst1ϕstm,nζ2p(F(|λxmn|ρ3))p+sups,t1supζUst1ϕstm,nζ2p(F(|μymn|ρ3))psups,t1supζUst1ϕstm,nζ(2p+1)(F(|xmn|ρ1))p+sups,t1supζUst1ϕstm,nζ(2p+1)(F(|ymn|ρ2))p<.

This shows that xy ∈ ℳ(ϕp, ℱ) ⇒ λxμy ∈ ℳ(ϕp, ℱ).

Remark 3.1

The distance measure between two sequences xn and yn induced by Cespq(F) can be represented as

dCespq(F)(x,y)=(n=1(1Qni=1nqifi(|xiyi|))p)1p.

Theorem 3.2

ℳ(ϕp, ℱ) ⊆ ℳ(ψp, ℱ) if and only if sups,t1(ϕstψst)<.

Proof

Let x ∈ ℳ(ϕp, ℱ). Then

sups,t1supζUst1ϕstm,nζ(F(|xmn|ρ))p<for some ρ>0.

Suppose that sups,t1(ϕstψst)<. Then φst ≤ kψst for some positive number k and for all st ∈ ℕ, so 1ψstkφst for all st ∈ ℕ. Therefore we have

1ψstm,nζ(F(|xmn|ρ))pkϕstm,nζ(F(|xmn|ρ))pfor each s,tN and for some ρ>0.

Now taking the supremum on both sides we get

sups,t1supζUst1ψstm,nζ(F(|xmn|ρ))pksups,t1supζUst1ϕstm,nζ(F(|xmn|ρ))pand for some ρ>0.

Therefore we have

sups,t1supζUst1ψstm,nζ(F(|xmn|ρ))p<and for some ρ>0.

Hence x ∈ ℳ(ψp, ℱ).

Conversely, let ℳ(ϕp, ℱ) ⊆ ℳ(ψp, ℱ) and suppose that sups,t1(ϕstψst)<. Then there exist increasing sequences (si) and (ti) of natural numbers such that lim(ϕstψst)=. Now for every b ∈ ℝ+, the set of positive real numbers, there exist ij ∈ ℕ such that φsitiψsiti>b for all si ≥ i and ti ≥ j. Hence 1ψsiti>bφsiti, so that, for some ρ > 0,

1ψsitim,nζ(F(|xmn|ρ))p>bϕsitim,nζ(F(|xmn|ρ))p

for all si ≥ i and ti ≥ j. Now taking the supremum over si ≥ i, ti ≥ j, and ζ ∈ 𝒰st, we get

supsii,tijsupζUst1ψsitim,nζ(F(|xmn|ρ))p>bsupsii,tijsupζUst1ϕsitim,nζ(F(|xmn|ρ))p.
2

Since (2) holds for all b ∈ ℝ+ (we may take b sufficiently large), we have

supsii,tijsupζUst1ψsitim,nζ(F(|xmn|ρ))p=

when x ∈ M(φpF) with 0<supsii,tijsupζUst1ϕsitim,nζ(F(|xmn|ρ))p<.

Therefore x ∉ ℳ(ψp, ℱ). This contradicts to ℳ(ϕp, ℱ) ⊆ ℳ(ψp, ℱ). Hence sups,t1(ϕstψst)<.

Corollary 3.1

ℳ(ϕp, ℱ) = ℳ(ψp, ℱ) if and only if sups,t≥1(ηst) < ∞ and sups,t1(ηst1)<, where ηst=(ϕstψst) for all st ∈ ℕ.

Corollary 3.2

ℳ(ϕ) ⊆ ℳ(ϕp, ℱ).

Proof

If p = 1 and ℱ(x) = x, then ℳ(ϕ) = ℳ(ϕp, ℱ). Also, ℳ(ϕ) ⊆ ℳ(ϕp, ℱ).

Theorem 3.3

The inclusions p(ℱ) ⊆ ℳ(ϕp, ℱ) ⊆ ℒ(ℱ)ℳ(ϕp, ℱ) hold.

Proof

Let x ∈ ℒp(ℱ). Then, for some ρ > 0, we have i,j=1,1,(F(|xmn|ρ))p<. Since (φst) is nondecreasing with respect to st ≥ 1, for some ρ > 0, we have

1ϕstm,nζ(F(|xmn|ρ))p1ϕ11m,nζ(F(|xmn|ρ))p1ϕ11i,j=1,1,(F(|xi,j|ρ))p<.

Hence sups,t1supζUst1ϕstm,nζ(F(|xmn|ρ))p<.

Thus p ⊆ ℳ(ϕp, ℱ). Now let x ∈ ℳ(ϕp, ℱ). Then for some ρ > 0, we have

sups,t1supζUst1ϕstm,nζ(F(|xmn|ρ))p<,supm,n11ϕ11m,nζ(F(|xmn|ρ))p<.

F(|xmn|ρ)(Aϕ11)1p for some A > 0 and all mn ∈ ℕ. Thus x ∈ ℒ(ℱ).

Theorem 3.4

Let , 1, 2 be Orlicz functions satisfying Δ2-condition. Then

  1. ℳ(ϕp, ℱ1) ⊆ ℳ(ϕp, ℱ ∘ ℱ1),
  2. ℳ(ϕp, ℱ1) ∩ ℳ(ϕp, ℱ2) = ℳ(ϕp, ℱ1 + ℱ2).

Proof

(1) Let x ∈ ℳ(ϕp, ℱ1). Then there exists ρ > 0 such that

sups,t1supζUst1ϕstm,nζ(F(|xmn|ρ))p<.

Let 0 < ε < 1 and δ with 0 < δ < 1 be such that F(t) < ε, 0 < t ≤ δ. Put tmn=F1(|xmn|ρ) and for any ζ ∈ 𝒰s, consider

m,nζ(F(tmn))p=1(F(tmn))p+2(F(tmn))p,

where the first sum is over tmn ≤ δ, and the second is over tmn > δ. From the remark we have

1(F(tmn))p(F(1))p1(tmnp)(F(2))p1(tmnp),
3

and for tmn > δ, we use the fact that

tmn<tmnδ<1+tmnδ.

Since is nondecreasing and convex, we have

F(tmn)F(1+tmnδ)<12F(2)+12F(2tmnδ).

Since satisfies Δ2-condition, we have

F(tmn)<12ktmnδF(2)+12ktmnδF(2)=ktmnδF(2).

Hence

(F(tmn))p<(ktmnδF(2))p.

Therefore

2(F(tmn))max(1,(kF(2)δ)p)pm,nζ(tmnp).
4

By (3) and (4) we have ℳ(ϕp, ℱ1) ⊆ ℳ(ϕp, ℱ ∘ ℱ1).

(2) The proof follows from the inequality

sups,t1supζUst1ϕst{m,nζ((F1+F2)(|xmn|ρ))p}1psups,t1supζUst1ϕst{m,nζ(F1(|xmn|ρ))p}1p+sups,t1supζUst1ϕst{m,nζ(F2(|xmn|ρ))p}1p<.

Theorem 3.5

The ℳ(ϕp, ℱ) satisfy the following relations:

  1. ℳ(ϕp, ℱ) = ℒp(ℱ) if and only if sups,t≥1(φst) < ∞,
  2. ℳ(ϕp, ℱ) = ℒ(ℱ) if and only if sups,t1(stϕst)<.

Proof

(1) If we take φst = 1 for all st ∈ ℕ, then we have ℳ(ϕp, ℱ) = ℒp(ℱ).

(2) By Theorem 3.3 we easily get that ℳ(ϕp, ℱ) = ℒ(ℱ) if and only if sups,t1(stϕst)<.

k-means algorithm for ℳ(ϕp, ℱ) distance measure

Let X = {x1x2, …, xn} be a given data set. Then the proposed clustering algoritm works as follows.

  1. Select first k data points as the cluster center xk = {x1x2, …, xk} (where k is the number of clusters).
  2. Compute the distance between each data point and cluster center through ℳ(ϕp, ℱ) distance measure.
  3. Put the data point into that cluster whose ℳ(ϕp, ℱ) distance with its center is minimal.
  4. Redefine cluster centers for newly evolved clusters due to the above steps; the new cluster centers are computed as ci=1kij=1kixi, where ki is the number of points in the ith cluster.
  5. Repeat Step 1 to Step 4 until the difference between two consecutive cluster centers becomes less than a desired small number.

Clustering by using the induced ℳ(ϕp, ℱ) distance measure

Two-moon and path-based data sets are artificially designed as nonconvex collections of points [36, 37]. The original shapes of the two-moon and path-based data are represented in Figures 2 and and4,4, respectively. The clustering on these two data sets is carried out by the algorithm dissussed in Section 3.1. In the case of a two-moon data set, for making simulation process simple, we take φ = 1, mn, p = 1, and F(x) = |x|. In Figure 3(a), it is shown that the clustering accuracy of the k-means clustering algorithm is 78% over the two-moon data set, whereas the clustering accuracy of our modefied algorithm k-means clustering is 84% (Figure 3(b)). Moreover, in the case of a path-based data set, we take φn, mn, p = 1, and F(x) = |x|. The clustering accuracy of the path-based data set by using the k-mean clustering algorithm is 45%, whereas by using the proposed modefied k-means clustering algorithm it is 67% as shown in Figures 5(a) and and5(b),5(b), respectively.

Figure 2
Two-moon data set.
Figure 3
Obtained clustering results for two-moon data set.
Figure 4
Path-based data set.
Figure 5
Obtained clustering results for two-moon data set.

Conclusions

The parameters ϕ, p, involved in the sequence space ℳ(ϕp, ℱ) give additional three degrees of freedom to its induced distance measure. Therefore, it is more flexible in comparison to the lp or weighted lp distance measure. The flexibility in the distance measure can be judiciously used in the clustering of the real-world data sets. We have proposed only a modified k-means clustering algorithm; in the similar fashion, other distance-based clustering algorithms can also be modified. So, improvement in many clustering algorithms is possible due to a distance measure of ℳ(ϕp, ℱ). We have shown the efficacy of an ℳ(ϕp, ℱ)-based k-means clustering algorithm over the l2-based k-means clustering algorithm on the basis of better clustering accuracy obtained for a two-moon data set and path-based data set.

Authors’ contributions

Authors’ contributions

Both authors of the manuscript have read and agreed to its content and are accountable for all aspects of the accuracy and integrity of the manuscript.

Notes

Competing interests

The authors declare that they have no competing interests.

Footnotes

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Contributor Information

Mohd Shoaib Khan, moc.liamg@umanahkbiaohs.

QM Danish Lohani, ni.ca.uas.sc@inaholhsinad.

References

1. MacQueen J. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability. 1967. Some methods for classification and analysis of multivariate observations; pp. 281–297.
2. Bezdek JC. A review of probabilistic, fuzzy, and neural models for pattern recognition. J. Intell. Fuzzy Syst. 1993;1(1):1–25. doi: 10.1109/TFUZZ.1993.6027269. [Cross Ref]
3. Khan VA, Lohani QM. New lacunary strong convergence difference sequence spaces defined by sequence of moduli. Kyungpook Math. J. 2006;46(4):591–595.
4. Jain AK. Data clustering: 50 years beyond k-means. Pattern Recognit. Lett. 2010;31(8):651–666. doi: 10.1016/j.patrec.2009.09.011. [Cross Ref]
5. Cheng MY, Huang KY, Chen HM. K-means particle swarm optimization with embedded chaotic search for solving multidimensional problems. Appl. Math. Comput. 2012;219(6):3091–3099.
6. Yao H, Duan Q, Li D, Wang J. An improved k-means clustering algorithm for fish image segmentation. Math. Comput. Model. 2013;58(3):790–798. doi: 10.1016/j.mcm.2012.12.025. [Cross Ref]
7. Cap M, Prez A, Lozano JA. An efficient approximation to the k-means clustering for massive data. Knowl.-Based Syst. 2017;117:56–69. doi: 10.1016/j.knosys.2016.06.031. [Cross Ref]
8. Güngör Z, Ünler A. K-harmonic means data clustering with simulated annealing heuristic. Appl. Math. Comput. 2007;184(2):199–209.
9. Kellogg CN. An extension of the Hausdorff-Young theorem. Mich. Math. J. 1971;18(2):121–127. doi: 10.1307/mmj/1029000635. [Cross Ref]
10. Jovanović I, Rakočević V. Multipliers of mixed-norm sequence spaces and measures of noncompactness. Publ. Inst. Math. 1994;56:61–68.
11. Blasco O, Zaragoza-Berzosa C. Multipliers on generalized mixed norm sequence spaces. Abstr. Appl. Anal. 2014;2014:983273. doi: 10.1155/2014/983273. [Cross Ref]
12. Dolović I, Malkowsky E, Petković K. New approach to some results related to mixed norm sequence spaces. Filomat. 2016;30(1):83–88. doi: 10.2298/FIL1601083D. [Cross Ref]
13. Sargent W. Some sequence spaces related to the lplp spaces. J. Lond. Math. Soc. 1960;1(2):161–171. doi: 10.1112/jlms/s1-35.2.161. [Cross Ref]
14. Tripathy BC, Sen M. On a new class of sequences related to the space lplp. Tamkang J. Math. 2002;33(2):167–171.
15. Mursaleen M. Some geometric properties of a sequence space related to pp. Bull. Aust. Math. Soc. 2003;67(2):343–347. doi: 10.1017/S0004972700033803. [Cross Ref]
16. Mursaleen M. Application of measure of noncompactness to infinite system of differential equations. Can. Math. Bull. 2013;56:388–394. doi: 10.4153/CMB-2011-170-7. [Cross Ref]
17. Khan VA. A new type of difference sequence spaces. Appl. Sci. 2010;12:102–108.
18. Malkowsky E, Savas E. Some λ-sequence spaces defined by a modulus. Arch. Math. 2000;36:219–228.
19. Djolović I, Malkowsky E. On matrix mappings into some strong Cesaro sequence spaces. Appl. Math. Comput. 2012;218(10):6155–6163.
20. Khan MS, Alamri BAS, Mursaleen M, Lohani QMD. Sequence spaces M(φ)M(φ) and N(φ)N(φ) with application in clustering. J. Inequal. Appl. 2017;2017(1):63. doi: 10.1186/s13660-017-1333-z. [PMC free article] [PubMed] [Cross Ref]
21. Mohiuddine SA, Srivastava HM, Alotaibi A. Application of measures of noncompactness to the infinite system of second-order differential equations in pp spaces. Adv. Differ. Equ. 2016;2016(1):317. doi: 10.1186/s13662-016-1016-y. [Cross Ref]
22. Alotaibi A, Mursaleen M, Mohiuddine SA. Application of measures of noncompactness to infinite system of linear equations in sequence spaces. Bull. Iran. Math. Soc. 2015;41(2):519–527.
23. Mursaleen M, Mohiuddine SA. Convergence Methods for Double Sequences and Applications. Berlin: Springer; 2014.
24. Altay B. On some double Cesàro sequence spaces. Filomat. 2015;28(7):1417–1424.
25. Altay B, Başar F. Some new spaces of double sequences. J. Math. Anal. Appl. 2005;309(1):70–90. doi: 10.1016/j.jmaa.2004.12.020. [Cross Ref]
26. Basar F, Sever Y. The space LqLq of double sequences. Math. J. Okayama Univ. 2009;51:149–157.
27. Esi A, Hazarika B. Some double sequence spaces of interval numbers defined by Orlicz function. J. Egypt. Math. Soc. 2014;22(3):424–427. doi: 10.1016/j.joems.2013.12.004. [Cross Ref]
28. Savas E, Patterson RF. On some double almost lacunary sequence spaces defined by Orlicz functions. Filomat. 2005;19:35–44. doi: 10.2298/FIL0519035S. [Cross Ref]
29. Mohiuddine SA, Raj K, Alotaibi A. Some paranormed double difference sequence spaces for Orlicz functions and bounded-regular matrices. Abstr. Appl. Anal. 2014;2014:419064.
30. Hazarika B, Kumar V, Lafuerza-Guillén B. Generalized ideal convergence in intuitionistic fuzzy normed linear spaces. Filomat. 2013;27(5):811–820. doi: 10.2298/FIL1305811H. [Cross Ref]
31. Talo O, Basar F. Certain spaces of sequences of fuzzy numbers defined by a modulus function. Demonstr. Math. 2010;43(1):139–149.
32. Mohiuddine SA, Hazarika B. Some classes of ideal convergent sequences and generalized difference matrix operator. Filomat. 2017;31(6):1827–1834. doi: 10.2298/FIL1706827M. [Cross Ref]
33. Mursaleen M, Sharma SK, Mohiuddine SA, Kılıçman A. New difference sequence spaces defined by Musielak-Orlicz function. Abstr. Appl. Anal. 2014;2014:691632.
34. Nakano H. Concave modulars. J. Math. Soc. Jpn. 1953;5:29–49. doi: 10.2969/jmsj/00510029. [Cross Ref]
35. Zeltser M. Investigation of Double Sequence Spaces by Soft and Hard Analitical Methods. Tartu: Tartu University Press, Univ. of Tartu, Faculty of Mathematics and Computer Science; 2001.
36. Jain AK, Law MHC. Proceedings of the First International Conference on Pattern Recognition and Machine Intelligence. 2005. Data clustering: a users dilemma.
37. Chang H, Yeung D-Y. Robust path-based spectral clustering. Pattern Recognit. 2008;41(1):191–203. doi: 10.1016/j.patcog.2007.04.010. [Cross Ref]
38. Khan MS, Lohani QMD. Computational Intelligence (IWCI), International Workshop on. Dhaka: IEEE; 2016. A similarity measure for atanassov intuitionistic fuzzy sets and its application to clustering.
39. Khan MS, Lohani QMD, Mursaleen M. A novel intuitionistic fuzzy similarity measure based on the double sequence by using modulus function with application in pattern recognition. Cogent Math. 2017;4(1):1385374.

Articles from Springer Open Choice are provided here courtesy of Springer