Composition Operations of Generalized Entropies Applied to the Study of Numbers

The generalized entropies of C. Tsallis and G. Kaniadakis have composition operations, which can be applied to the study of numbers. Here we will discuss these composition rules and use them to study some famous sequences of numbers (Mersenne, Fermat, Cullen, Woodall and Thabit numbers). We will also consider the sequence of the repunits, which can be seen as a specific case of q-integers.


Introduction
In some recent works [1][2][3], we have discussed the generalized entropies of C. Tsallis [4] and G. Kaniadakis [5,6], with the aim of applying them to image processing and image segmentation. What is quite attractive of these entropies is the fact that they are non-additive. It means that they are following rules of composition, which are different from the usual operation of addition. Moreover, these composition rules contain indices, which are useful to have a specific segmentation of images, or even to drive a gray-level image transition among the textures of the image [7].
The Tsallis composition rule is defined in [8] as a pseudo-additivity. However, as these rules are concerning generalized entropies, we could call them "generalized sums". Actually, I used this locution in a discussion about the rules of composition that we can obtain from the transcendental functions [9]. The approach, given in [9], is using a method based on the generators of algebras [10][11][12]. Here I show that we can use the generalized sums, as those that we can obtain from Tsallis and Kaniadakis generalized statistics, to the study of the sequences of numbers. In particular, we will discuss the composition rules that can be applied to famous sequences of numbers, such as Fermat, Mersenne and Thabit numbers. We will also consider the sequence of the repunits, which are a specific case of q-integers.
Let us start remembering the composition rules of Kaniadakis and Tsallis entropies.

The generalized sum of Kaniadakis statistics
In [6], a generalized sum is defined in the following manner. Let us consider two elements x and y of reals R, and a parameter real too. The composition law is given by: which defines a generalized sum, named -sum. Reals R and operation (1) form an Abelian group.
Let us remember that a group is a set A and an operation •. The operation combines any two elements a,b to form another element of the group denoted a•b.
To qualify (A,•) as a group, the set and operation must satisfy the following requirements where e is the identity (the notation is inherited from the multiplicative operation).
If a group is Abelian, a further requirement is the commutativity. For all a,b in A, a•b = b•a. Therefore, to qualify a group as an Abelian group, the set and operation must satisfy five requirements, which are known as the Abelian group axioms. A group having a not commutative operation is a "non-Abelian group" or a "non-commutative group". For an Abelian group, one may choose to denote the group operation by + and the identity element by 0 (neutral element) and the inverse element of as (opposite element). In this case, the group is called an additive group.

Amelia Carolina Sparavigna 1
We can obtain the operation of a group by means of functions. Actually, if a function ( ) exists, which is invertible ( ( )) , we can use it as a generator, to generate an algebra [10]. In [11], G is used to define the group law ( ), such as: In the case of the -sum, the function G is the hyperbolic sine, so that [9]: This operation is used in Kaniadakis generalized statistics for the sum of relativistic momenta. A property of the hyperbolic sine function is that: The domain is the whole real line. We have [13]: Therefore, we have that the group law is given as [9]: . Therefore, the generalized sum is: ) and ( ( )); we can easily see that [9]:

Other groups
Of course, other generalized sums can be obtained. Let us consider, for instance, the hyperbolic tangent. Its inverse hyperbolic function is defined as: The domain is the open interval ( ). A property of this function is the following [13]: We have a group law given by ( ) ( ( ) ( )). Therefore, we can obtain the generalized sum as [9]: This sum is commutative. The neutral element is 0.
The opposite element of x is . We have the associativity, that is ( ) ( ) [9]. To show that we have a group of (-1,1) and operation (3) it is necessary to verify the closure. That is, we have to see that the result of the sum is in the open interval (-1,1). This is discussed in [9]. The generalized sum (3) is used by Kaniadakis in Ref.14 for the relativistic velocity. Another group is obtained from function ( ) ( ) and its inverse ( ) ( √ ).
The group law ( ) = ( ( ) ( )) gives [9]: We will show in a following section that the same rule of additivity exists for the Fermat numbers.

The q-integers
Let us see how we can apply the previously discussed generalized sums to the numbers. Let us start from the q-integers of the q-calculus.
Many mathematicians have contributed to this calculus [15][16][17][18][19][20]. Consequently, the q-calculus is also known as "quantum calculus" and "time-scale calculus", or "calculus of partitions" too [19]. Moreover, it is expressed by means of different notations or, as told in [19], by different "dialects". Here we will use the notation given in the book by Kac and Cheung [20].
As discussed in [21], the q-integers are forming a group having a generalized sum, which is similar to sum of the Tsallis q-entropy of two independent systems. The symmetric q-integers are linked to the Kaniadakis calculus.
First, we have to define the operation of addition by composing two q-integers. This operation is not the sum that we use for the natural integers of course, but it is a generalized sum.
Let us start from the q-integer [ ] and calculate as in [21]. We have: This is the same as (1), for parameter = 1. The closure is given by the fact that the result of this operation is on the real line. The neutral element is 0. The opposite element of x is . Also the commutativity is evident.
To have a group, we need the discussion of the associativity too, showing therefore that ( ) ( The generalized sum (5) is associative [21]. We have also that: Therefore, the five axioms of an Abelian group are satisfied. In this manner, using the generalized sum given by (5), we have the Abelian group of the qintegers.
What is important for the present discussion is the fact that the generalized sum (5) is similar to the sum that we find in the approach to entropy proposed by Constantino Tsallis [4], for his generalized entropy. For two independent systems A and B, the Tsallis entropy is given by: In this formula the parameter ( ), in a certain manner, measures the departure from the ordinary additivity, which is recovered in the limit . The group on which is based the Tsallis entropy, and therefore the generalized sum given above, is known in literature as the "multiplicative group". As told in [12], the use of a group structure allows determining a class of generalized entropies.

Symmetric q-numbers
In the previous section, we have considered the group of the q-integers as defined by q-calculus. In [20], it is also defined the symmetric q-integer in the following form (here we use a notation different from that given in the Ref. 20):

[ ]
Repeating the approach previously given, we can determine the group of the symmetric q-integers. Let us start from the q-integer[ ] , which is: and try to find it as a generalized sum of the qintegers [ ] and [ ] [21].
By writing ( ), the q-integer turns out into a hyperbolic sine: from (6), after some passages using (7), we find [21]: ). Therefore, we have the generalized sum of the symmetric q-integers as: Let us stress that (8) is also the generalized sum (1) proposed by G. Kaniadakis (see also the discussion in [22]).

The Mersenne numbers
In the case that , we have: These are the Mersenne Numbers. About these numbers, a large literature exists (see for instance that given in [23]). Among the Mersenne integers, we find the Mersenne primes. The numbers are named after Marin Mersenne (1588 -1648), a French Minim friar, who studied them in the early 17th century. Let us call the Mersenne number. Of course, we have the generalized sum for the qnumbers as given in (5):

[ ] [ ] [ ] [ ] [ ][ ]
However, we can repeat the calculus starting from . After some passages we obtain = ( )( ) , that is: (9) Let us stress that we have a generalized sum of the form of those of the groups known as "multiplicative groups". Using (9), for the Mersenne numbers we can imagine the following recursive relation: The sum (9) is associative. We cannot have a group of the Mersenne numbers, without considering also the opposites O of them, so that: ( ). Therefore:

( )
Explicitly: ( ) = . These numbers are the Mersenne numbers with a negative exponent.

Symmetric Mersenne
Let us consider again the symmetric q-integer in the case of . We can define the symmetric Mersenne in the following manner: By writing ( ), (10) turns out into a hyperbolic sine: Let us define ( ) ⁄ . We have the generalized sum of the symmetric Mersenne as: Of course, we have again the generalized sum proposed by G. Kaniadakis.

Fermat, Cullen and Woodall Numbers
As seen before, the Mersenne numbers are forming a group with the following generalized sum: Using this composition rule, we can have the following recursive relation: Another famous sequence of integers is that of the Fermat numbers. Fermat numbers are defined as [24]. These numbers have the following generalized sum [25]: which is the same of the group law (4) that we have discussed in [25] and in a previous section of this work.
The sum is associative. The neutral element is and the opposites O of the Fermat numbers are given by ( ) [25].
Similar to the Fermat numbers, we have the Cullen numbers. The Woodall numbers are similar to the Mersenne numbers [26,27]. Let us find the generalized sums of them, as detailed in [28].
The Cullen numbers are . The generalized sum is [28]: The neutral element of this sum is . Using (13) Of the Cullen numbers, we can also give another form of the generalized sum [28]: The recursive relation assumes the form: In the case that we use the generalized sum (14), we have to remember that when m or n are equal to zero, we need to assume ( ) ⁄ , ( ) ⁄ . The Woodall numbers are defined as: . Let us write the generalized sum as [28]: The neutral element of this sum is . Using (15) (15), as shown in [28].

Thabit numbers
Let us consider the Thabit numbers [29]. These numbers are given as , where the asterisk represents the ordinary multiplication. The operation of addition between Thabit numbers is [30]: Using (16), we can see that the neutral element is , so that: ( ) The recursive relation is given accordingly to (

[ ]
We can use the same approach for the repunits of that given for the q-numbers. Let us consider the following operation (generalized sum): . It is defined in the following manner: The recursive relation for the repunits, given according to (17) and starting from , is:

Conclusion
In this work, we have discussed the composition operations of generalized entropies (Tsallis and Kaniadakis). These operations can be obtained from some group laws based on functions and their inverses. The group laws can be defined as "generalized sums" because the generalized entropies are motivating them.
The approach using the group operations can be applied to the study of numbers. We have discussed the composition rules for some famous sequences of numbers (Mersenne, Fermat, Cullen, Woodall and Thabit numbers). We have also considered the sequence of the repunits, which can be seen as a specific case of the q-integers.