I am trying to find if there is a mathematical formula that can, for example, convert a function f(x,y) to f(r) such that f(x0,y0) = f(r0) and f(x+a,y+a) = f(r+a).
Yes, you can. However in spacial cases. For instance if f(x,y) obeys scaling low.
In priciples you can assume that it does and then you have to make an experimental test. f(x,y) obeys the scaling low if this is homogenous function in general sense, which means that there exists real a,b,c such that for each lambda>0 the following relation holds: f(lambdaa x,lambdab y)=lambdac f(x,y). According to assumptions we are free to substitute: lambda=x-1/a. Then we obtain:
f(1,x^(-b/a)y)=x^(-c/a)f(x,y). On the left hand side you have got function of the one variable z=x^(-b/a)y.
Thank you so much...this is what I was looking for..
What is the name of this theory and is there a reference book? Because I want a more generalized case i.e f(x1,...,xn) converted to f(y1,...yp) such that p
In some cases you can use the polar coordinates, where r=sqrt(x^2+y^2), but you need also a kind of symmetry in order to get rid of polar angle theta...
Name of theory is SCALING. By the scaling one understands an invariance of analytic properties with respwct to transformation which I have described in my previous message. Book by G.I Barenblat : "SCALING", Cambridge Texts in Applied Mathematics., Cambridge University Press. I suggest also book by E. Stanley, "Phase transitions and Critical Phenomena. Do the best.
It's possible that I haven't understood exactly what you want to accomplish, but it is clear that in general the answer to your question is "no". If it were possible to reduce the number of variables always, by some method, then there would hardly be any need to study functions of more than one variable.
Yes, you can repeat the scaling algoritm for the multi variables function:
f(lambda^a x, lambda^b y, lambda^c z.....)=lambda^k f(x,y,z,.....). Let us drop out the second argument: lambda=y^(-1/b), -> f(y^(-a/b) x, 1, y^(-c/b) z, ...)=y^(-k/b) f(x,y,z,...).
You always can but it is not free and you cannot do it keeping invariance under addition: There is a surjective continuous map
H:[0,1] \to [0,1] \times [0,1]
the so called Hilbert plane filling curve. That map is not differentiable and it is not one to one. Therefore a function f(x,y) on [0,1] \times [0,1] is completely determined by (f\circ H)(t) on [0,1].
However if you want to replace a differentiable function with a differentiable function of one variable less, you really do need some symmetry, like homogeneity (f effectively depends only on the angle with the x axis) or rotational invariance (f effectively depends only on the radius).
Of course, one can always "reduce" a function of several arguments to a function of a single argument, since product and Hom are adjoint functors, that is, since there is a natural transformation between
hom(A x B, C)
and
hom(A, C^B)
This general property was known to Frege and Schoenfinkel. My answer is really a response to the statement made by prior to mine, regarding the statement
"If it were possible to reduce the number of variables always, by some method, then there would hardly be any need to study functions of more than one variable."
I think in some cases, it is possible but if we could do this in general case, the theory of n variable functions was redundant. As I know, for some special positive n-linear map on C(K) spaces where K is a compact Hausdorff space, there is a way to reduce the variable. Please see my joint paper with Vladimir Troitsky : doi: 10.1007_s11117-013-0239-3, Theorem 1.
Rogier Brussee and Mycroft Holmes , can you please suggest some text books or papers that talks about these theories when used to reduce the number of variables of a function? I would like to read in depth about these theories.
The Wikipedia articles https://en.wikipedia.org/w/index.php?title=Hilbert_curve&oldid=633637210 and https://en.wikipedia.org/w/index.php?title=Space-filling_curve&oldid=624328726 are actually quite good for a practical and theoretical overview.
Mycroft Holmes trick is called adjunction in mathematics and currying in computer science. It is a very simple trick:
given f: X \times Y \to Z
we have for every x \in X a function
F_x : Y \to Z ,
F_x: y \to f(x,y)
and so we have map
\phi_f : X \to Map(Y,Z),
x \to F_x
Note that unlike the Hilbertcurve construction this construction is natural in the sense that it is unique and behaves well with respect to maps X\times Y \to X' \times Y'
This is the version for the category of sets. Things get more interesting if we work in different categories, because we now have to define "products" and "Hom objects" in the category. (in most categories maps are called morphisms or homomorphisms) . Two examples:
In finite dimensional vectorspaces we have a "product" of two vectorspaces, the tensor product V\tensor W[1] and Hom object Hom(V,W) is just the vectorspace of linear maps from V to W. We have the adjunction map:
LInmaps(U \tensor V, W) \iso Linmaps( U, Hom(V,W))
In the category of topological spaces we have a product X \times Y of two topological spaces X and Y. We also have the space of continuous maps Map(X,Y) where the topology is defined by defining an open set as unions of finite intersections of open sets V(K,U) consisting of all maps that map a compact set K in an open set U. For locally compact spaces X,Y,Z we then have
ContinuousMaps(X \times Y, Z) \iso ContinuousMaps(X, Map(Y,Z))