Abstract: The aim of this article is to generalize the best approximation problem to the case of quotient generalized 2-normed spaces. At first, a quotient 2-norm is introduced and with an example we show that the quotient 2-norm is a generalized 2-norm that it is not a 2-norm. Afterward some theorems of approximation in quotient spaces are extended to this newly introduced quotient generalized 2-normed space in order to assess their validity. 2000 Mathematics Subject Classification. Primary 46A12.
INTRODUCTION
Approximation theory has many important applications in various areas of functional analysis, computer science, numerical solutions of differential and integral equations. As a generalization of normed spaces is 2-normed spaces that play a very important role in functional analysis. What we offer in this study is to investigate approximation theory in generalized 2-normed spaces. The concept of linear 2-normed spaces was introduced by Gahler (1964) as an interesting non-linear generalization of a normed linear space which was developed extensively in different subjects by others. During 1999-2006 (Lewandowska, 1999, 2001, 2003a, 2003b, 2004; Lewandowska et al., 2006) has published a series of papers on 2-normed sets and generalized 2-normed spaces. Lal and Das (1982), Chen (2002) and Rezapour (2009) carry on the development of this concept to 2-functional and approximation in 2-normed spaces.
Let X be a linear space of dimension greater than 1 over K, where K is the field of real or complex numbers. Suppose ||.,.|| be a nonnegative real-valued function on (XxX) satisfying the following conditions:
• | ||x, y|| = 0 if and only if x and y are linearly dependent vectors |
• | ||x, y|| = ||y, x|| for all x, yεX |
• | ||λx, y||= |λ|||x, y|| for all λεR and x, yεX |
• | ||x + y, z||≤||x, z||+||y, z|| for all x, y, zεX |
Then ||.,.|| is called a 2-norm on X and (X, ||.,.||) is called a linear 2-normed space. Every 2-normed space is a locally convex topological vector space. In fact for a fixed bεX, pb (x) = ||x, b|| is a seminorm and the family P = pb: bεX of seminorms generates a locally convex topology on X.
Definition 1: Let X and Y be linear spaces, D be a nonempty subset of XxY such that for every xεX and yεY, the sets:
are linear subspaces of the spaces Y and X, respectively. A function ||.,.||: D→[0, ∞) is called a generalized 2-norm on D if it satisfies the following conditions:
• | (N1) ||αx, y|| = |α|||x, y|| = ||x, αy|| for all (x, y)εD and every scalar α |
• | (N2) ||x, y + z||≤||x, y|| + ||x, z|| for all (x, y), (x, z)εD |
• | (N3) ||x, y + z||≤||x, z|| + ||y, z|| for all (x, z), (y, z)εD |
Then (D, ||.,.||) is called a 2-normed set. In particular, if D = XxY, (XxY, ||.,.||) is called a generalized 2-normed space. Moreover, if X = Y, then generalized 2-normed space is denoted by (X, ||.,.||).
Definition 2: Let X be a real linear space. Denote by X a non empty subset XxX with the property X = X-1 and such that the set Xy = {xεX; (x; y)εX} is a linear subspace of X, for all yεX. A function ||.,.||: X→[0, ∞) satisfying the following conditions:
• | (S1) ||x, y|| = ||y, x|| for all (x; y)εX |
• | (S2) ||αx, y|| = |α|||x, y|| = ||x, αy|| for any real number α and all (x, y)εX |
• | ||x, y + z||≤||x, y|| + ||x, z|| for all x, y, zεX such that (x, y), (x, z)εX |
will be called a generalized symmetric 2-norm on X. The set X is called a symmetric 2-normed set. In particular, if X =XxX, the function ||.,.|| will be called a generalized symmetric 2-norm on and the pair (X; ||.,.||) a generalized symmetric 2-normed space.
Every 2-normed space is a generalized 2-normed space but the converse is not true. The following example is a generalized 2-normed space that is not a 2-normed space.
Example 1: Let X be a real linear space having two norms ||.||1 and ||.||2. Then (X, ||.,.||) is a generalized 2-normed space with the 2-norm defined by:
Specially if ||.||1 = ||.||2, our generalized 2-normed space will be a generalized symmetric 2-normed space.
For further examples see (Lewandowska, 1999, 2001, 2003a, 2003b, 2004; Lewandowska et al., 2006).
BEST APPROXIMATION IN QUOTIENT GENERALIZED 2-NORMED SPACE
In the following theorem a generalized 2-norm on the space:
is defined.
Theorem 1: Let (XxY, ||.,.||) be a generalized 2-normed linear space and G1 and G2 be subspace of X and Y, respectively. Define:
for every xεX and yεY. Then |||.,.||| is a generalized 2-norm on.
Proof: First we prove that |||.,.||| is well defined.
If x1 +G1 = x2 + G2 and y1 + G2 = y2+G2, then there exists (g, g)εG1xG2 such that x1 = x2 + g and y1 = y2 + g. So:
Step 1:
Step 2:
Step 3:
Step 4: Similar to Step 3 we can show that:
Example 2: Let X = Y = R2 , G1 = x axis and G2 = y axis and let:
where, x = (x1, x2), y = (y1, y2). It is a generalized 2-norm on R2. Then:
It is obvious that |||.,.||| is a generalized 2-norm that it is not a 2-norm.
Definition 3: Let (XxY, ||.,.||) be a generalized 2-normed space, G1 be a subspace of X and let G2 be a subspace of Y . Then, G1xG2 is called 2-proximinal if for every (x, y)ε XxY there exists (g0, g0) such that:
In this case, (g0, g0) is called 2-best approximation
of (x, y) in G1xG2 and the set of all 2-best approximations
of (x, y) in G1xG2 is denoted by
Theorem 2: Let (XxY, ||.,.||) be a generalized 2-normed linear space, K1, G1 and K2, G2 be subspaces of X and Y, respectively such that K1⊂G1 and K2⊂G2. Then the following are true:
(1) |
If (g1, g2) is a 2-best approximation
to (x, y) in G1xG2, then (g1 + K1,
g2 + K2) is a 2-best approximation to (x + K1,
y + K2) in |
(2) | If (g1 + K1, g2 + K2) is a
2-best approximation to (x + K1, y + K2) in |
(3) | If K1xK2 is 2-proximinal in XxY and |
(4) | If K1xK2 is a 2-semi-chebyshev subspace in XxY and
|
(5) | If G1xG2 is 2-proximinal in XxY, then |
(6) | If K1xK2 be 2-chebyshev in XxY and |
(7) | If K1xK2 is 2-proximinal in XxY and G1xG2
is 2-semi-chebyshev in XxY, then |
(8) | If K1xK2 is 2-proximinal in XxY and G1xG2
is 2-chebyshev in XxY, then |
Proof
Step 1: If (g1 + K1, g2 + K2)
is not a 2-best approximation to (x + K1, y + K2) in
Hence, there exists (g0, g0)εG1xG2 such that |||x-g0+K1, y-g0 + K2|||<|||x-g1 + K1, y-g2+K2|||. Since |||x-g1+K1, y-g2+K2|||≤||x-g1, y-g2, we have |||x-g0+K1, y-g0+K2|||<||x-g1, y-g2||. Thus, for some (k1, k2)εK1xK2, |||x-g0+k1, y-g0+k2|||<||x-g1, y-g2||. K1⊂G1 and K2⊂G2 implies that (g0+k1, g0+k2)εG1xG2.Therefore, (g1, g2) is not a 2-best approximation to (x, y) in G1xG2 and this is a contradiction.
Step 2: By hypothesis:
and |||x-g1+K1, y-g2+K2|||≤|||x-g1+K1, y-g2+K2||| for all (g1, g2)εG1xG2. Therefore for any (g1, g2)εG1xG2:
Therefore:
Step 3: It is clear by step 2.
Step 4: Let the conclusion be false, then there exists (x, y)εXxY
that has two distinct-best approximation (g1, g2) and
(g3, g4) from (G1xG2). By Step 1,
(g1+K1, g2+K2) and (g3+K1,
g4+K2) are 2-best approximations to (x+G1,
y+G2) from
Therefore (k1, k2) and (0, 0) are 2-best approximations to (x-g1, y-g2) from (K1xK2). Since (k1, k2)≠(0, 0), then (K1xK2) is not 2-semi-chebyshev in XxY and this is a contradiction.
Step 5: It is an immediate conclusion from step 1.
Step 6: It is clear by step 3 and 1.
Step 7: If it is not, then there exists
Step 7: It is clear by step 6 and 7.
Let G1xG2 be a 2-proximinal subspace of a generalized 2-normed linear space XxY and let denote the quotient space equipped with the generalized 2-norm |||.,.||| defined above. Then we define the quotient map:
by π (x, y) = (x+G1, y+G2).
Theorem 3: Let G1xG2 and K1xK2 be subspaces of a generalized 2-normed linear space XxY and K1xK2. If K1xK2⊂G1xG2 be 2-proximinal in G1xG2, then:
Furthermore if K1xK2 be 2-proximinal in XxY, then:
Proof: By part (1) of theorem (2), it is clear that:
Now let K1xK2 be 2-proximinal in XxY. By part (2) of theorem (2), if:
and:
then:
Therefore:
Hence,
Theorem 4: For a linear subspace G1xG2 of a generalized 2-normed linear space XxY, the following statements are equivalent:
(1) |
G1xG2 is proximinal |
(2) | We have: |
(3) | G1xG2 is closed and for the canonical mapping |
We have:
Proof: (1)⇔(2) If G1xG2 is 2-proximinal,
(x, y)εXxY (G1xG2) and
conversely, if we have (2) and (x, y)εXxY then (x, y) = (g1+x1,
g2+y1), where, (g1, g2)εG1xG2
and
Hence:
(1)⇔(3) If G1xG2 be 2-proximinal and:
Then:
and π (x-g1, y-g2) = (x+G1, y+G2). Conversely, if we have (3) and (x, y)εXxY, then:
So (x+G1, y+G2) = π (x1, y1),
where
So
CONCLUSION
In this study we introduce the concept of best approximation in generalized quotient 2-normed spaces and present some results.