Abstract
We introduce notions of continuous strong Weihrauch reducibility and of continuous Weihrauch reducibility for functions with range in a preordered set. Then we associate with such functions certain labeled forests and trees and show that Wadge reducibility, continuous strong Weihrauch reducibility and continuous Weihrauch reducibility between such functions can be characterized by suitable reducibility relations between the associated forests when they are defined. This leads to a combinatorial description of the initial segments of these three hierarchies for those functions defined on the Baire space that have countable range in a bqo set and that are
Keywords
Introduction
Reducibility relations are natural tools for comparing computational problems with respect to their computational complexity. In this article three topological reducibility relations are considered. They are
Wadge reducibility, continuous strong Weihrauch reducibility, and continuous Weihrauch reducibility
Wadge reducibility was introduced by Wadge in his thesis [24] for subsets of the Baire space
In his PhD thesis [6–8] the author considered functions that are defined on a subset of the Baire space and that have finite discrete range (one may call them partial k-partitions if k is an upper bound for the cardinality of the range) and described the finite initial segments of the three continuous reducibility hierarchies of such functions. For the description he used certain trees that describe the discontinuities of such functions and three suitable reducibility relations between forests consisting of such trees. The result for Wadge reducibility of partial k-partitions was extended by Selivanov first to all
In this article the results from [6–8] are extended to
Furthermore, we show that using the theory of admissible representations one can transfer these observations to relations (also known as multi-valued functions)
In the following sections we introduce some notation and some notions that we will need, many of them well-known, and some of them perhaps not that well-known. Then, in Section 4 we give general definitions of Wadge reducibility, of continuous strong Weihrauch reducibility and of continuous Weihrauch reducibility for relations
Sets, relations, and functions
Sets
For a set X we denote by
Relations and functions
Let X and Y be sets. Then
The relation
If S is a subset of X then the restriction
A relation
Finite and infinite sequences
Let
Relations with special properties
Let X be a set. A relation
reflexive if, for all
transitive if, for all
preorder or quasi-order if it is reflexive and transitive,
symmetric if, for all
equivalence relation if it is a preorder and symmetric,
anti-symmetric if, for all
partial order if it is a preorder and anti-symmetric.
total order or linear order if it is a partial order and, for all
If ⩽ is a preorder on a set S, then usually we use infix notation, that is, we write
For any set X the prefix relation ⊑ on
If ⩽ is a preorder on a set X, then we define the relation ≡ on X as usual by
A pair
An element x of a preordered set
Let X be a set, and let ⩽ be a relation on X. An element
Let
For a set X with a relation ⩽ on X and for an element
Let
We call a preorder ⩽ on a set X and the pair If X is the empty set then there is only one relation ⪯ on X: the empty relation
Let X, Y, Z be sets. The composition of two relations
The composition operation ∘ is associative: For sets W, X, Y, Z and relations
This is well known. We omit the proof. Due to the associativity of ∘ one may omit brackets in repeated applications of ∘. Note that if
The product of two relations
Let X, Y, Z,
We omit the proof. For any set X we define the total function
For any relation
For any function
We omit the proof.
A relation
Let X,
Let X, Y, Z be sets and
We observe
We have
□
Relations can be used in order to describe computation problems. For example, given some input x from some set X, one may wish to compute some y from some set Y such that
If a function
The following lemma will be used later.
Let
If f is injective then
The relation
If f is surjective then
We have This is clear. We have This follows directly from the previous assertion.
□
Topological spaces
For basic topological notions the reader is referred to [11]. The following two examples of topologies will be used later.
For any set X, the set Let
A topology τ is called second-countable if there exists a countable basis for it. A topological space
Let Y be a topological space as well. A function
Let X be a metrizable space. For all countable ordinals
Let
We define the total bijection
Admissible representations
Admissible representations were introduced by Kreitz and Weihrauch [13] for second countable
If X is a set and
Let X be a set and let ρ and σ be representations of X. We say that ρ is reducible to σ and write Two representations ρ and σ of a set X are called equivalent (and then we write If X is a topological space, then a representation
It is clear that the relation ⩽ on representations of a fixed set X is a preorder. Hence, the relation ≡ on representations of a fixed set X is an equivalence relation. If for a topological space X there exists an admissible representation then all admissible representations of X are equivalent.
Let
If X is a set and
In particular, all admissible representations of a topological space induce the same topology on the space. Kreitz and Weihrauch [13] showed that for every second countable
Let X, Y be topological spaces, and let
If
If
Let
For a function
The function F is a refinement of the relation
The function
For all
The proof is straightforward. We omit it.
A function The relation R is called
The following lemmas are well known.
Let
If X and Y are sets, ρ and
Let X and Y be
Let X, Y, Z be
Wadge reducibility of relations
In this section, X,
Let X and
Note that
Let
There exists a continuous function
There exists a
Concerning Condition (4), note that
In the rest of the proof we assume that
Let X,
It is straightforward to see that the relation
Let Y be an arbitrary set. For two relations
This follows from the equivalence of the first and the fourth of the conditions in the previous lemma because the identity function
Let X, Let We write
Note that
Let
There exist continuous functions
There exist a
Concerning Condition (3) note that
Let X,
It is straightforward to check that
Let X,
Note that
The equivalence of the second and the third condition in the following lemma is a topological version of a computability-theoretic equivalence observed by Brattka, Gherardi and Pauly [2, Proposition 3.2.1]; compare also Gherardi and Marcone [4]. We omit the proof of the lemma.
Let
There exist continuous functions
There exist a
Let X,
It is straightforward to check that
Finally, let us consider the special case when both Y and
Let X,
First, we define admissible representations of the discrete spaces Y and
We are going to show that the second condition, Condition (2), and Condition (3) are equivalent. Let
For any natural number
Of course, this implies
Let us prove the first claim directly. Let us fix some
Let
Of course, this implies
Let us prove the first claim directly. Let us fix some
Let X and
Wadge reducibility
Let
It is clear that
Let Y be a set, and let the relation ⪯ on Y be simply the equality relation. Then for any topological spaces X,
The proof is straightforward, and we omit it. In fact, with
Let Y be a set, let X,
Here, the second condition is based on the functions
Let X,
By Lemma 4.2 the condition
Let
For functions
This obviously extends the Wadge reducibility introduced in Section 5.1 as, in case
Let X,
The proof is straightforward. We omit it.
For sets Y,
Let Y,
Here, the condition on the right hand side is based on the functions
Indeed:
Let us consider
Let X,
There exists a relatively continuous relation
By definition, the condition
We pay special attention to the case of relations with discrete range.
Let X,
There exists a good function According to Corollary 5.8 the first condition, There exists a relatively continuous relation “ “
Therefore, it is sufficient to show that the second condition above and this condition are equivalent.
Let
We call a function for all for all
For a good function For functions
Let
This is obvious. We omit the proof. Next, we show that the relation
Let X,
For any good function
For any good function
In the case
Let Let Let us assume
□
Let X,
If
If
If
Due to Lemma 5.13.3 the second and the third assertion are special cases of the first assertion.
Let us prove the first assertion. Let
The relation
The transitivity follows from Lemma 5.14.1. The reflexivity follows from the reflexivity of
Using
Let X,
It is sufficient to observe that a function
In fact, with
Let X,
Here, the second condition is based on the functions “ “
A relation ⪯ on a set Y is called a well-quasi-order if it is a preorder and for every infinite sequence
For any infinite set
Concerning the considered Y-arrays, this notion is surprisingly stable. We just remark that one arrives at the same notion if on
We will need the following assertions. All of them are well known; see, e.g., [14,21].
Let Y be a set, and let ⪯ and If If
Every finite preorder is a bqo.
Every bqo is a well-quasi-order.
For a preordered set
If
In this section we introduce labeled trees and relations
A reducibility on labeled forests corresponding to Wadge reducibility on functions
In descriptive set theory and set theory there are various kinds of definitions of trees; see e.g. [3,11]. We will need only two special classes of trees and will use a terminology convenient for our needs. Remember that we call a subset
A forest is a subset S of A tree is a forest S that has a ⊑-smallest element. This will be called the root of the forest and will be written
Note that we do not demand that the set S is a ⊑-downwards closed set. One could demand that, but it would make the proofs of some lemmas later on more cumbersome. If S is a subset of
Let Y be a set. A labeled forest (resp. tree) with labels in Y is a pair
Let Let We define a relation
It is obvious that
Let
If M is empty then the
Let
We choose a forest
Let
Let
if
if
We omit the obvious proof. This lemma allows us to define a function
In the following proof we use the following notation. For a labeled forest
Let
Let us fix some tree
The following proposition is a corollary of a stronger result by Laver [14].
Let
Let
This follows from the previous proposition and from Lemma 6.1.3. □
In this section
Let
Let We define a relation
First, we observe that the relation
Let
This follows directly from the definitions. □
Next, we formulate some properties of the relation
Let
This is obvious. We omit the proof.
Let
If
If
If
Due to Lemma 7.11 the second and the third assertion are special cases of the first assertion.
Let us prove the first assertion. Let
The relation
The transitivity follows from Lemma 7.13.1. The reflexivity follows from the reflexivity of
Let
If
The first assertion follows from Lemma 7.11 and from Lemma 7.13. The second assertion follows directly from the first assertion. □
Hence, for any good function
Let
We choose a tree
Let
This is obvious from the construction of a forest in
The following technical statement will be used in several proofs later on.
Let
there exists a good extension
and in the case
Let us fix a forest
For the second assertion we just need to remember that
Finally, we show that in special cases the relation
Let
There exists a monotone function
The previous lemma applies in particular to the following two cases.
When the set Y is finite and the relation ⪯ is the equality relation on Y. If When there is some finite set Z with
In the first of the two cases considered in Remark 7.20 the relation Let Y be a finite set and let ⪯ be the equality relation on Y. Let
There exists a monotone function
The implication “ In the case, when Y and
In this section
Let We define a relation
First, we show that the relation
Let
In the case
For any good function
For any good function
This is clear. Let Let
Let
Let For reflexivity consider
□
Note that the first assertion shows that
Let
If
The first assertion follows from Lemma 7.24.1 and from Lemma 7.25.1. The second assertion follows directly from the first assertion. □
Hence, for any good function
Let
The proof is almost identical with the proof of Lemma 7.16. □
Let
This is obvious from the construction of a forest in
Let
The proof is similar to and easier than the proof of Lemma 7.18. □
Finally we show that in special cases the relation
Let
There exists a monotone function
The previous lemma applies in particular to the two cases exhibited in Remark 7.20. In the case, when Y and
The definition of the trees and forests
Let
If
For any countable ordinal α and any
The following lemma justifies this definition.
Let α be a countable ordinal.
If
If
If
If
If, for some
We show this by induction over α. Let Let us assume that If Let us assume that some open subset If Let us assume that
□
In general,
The forest class
For all
The first assertion follows from the fact that set
Let
for every
for every
We show this by induction over α. For By induction hypothesis Let us fix some The second assertion implies that
□
We remark that later, in Proposition 8.13, we shall formulate an interesting strengthening of the second assertion of Proposition 8.3.
Let us compute the forests describing the discontinuity of the relations A representant of 
In Example 4.13 we introduced a relation

A representant of
Let

A representant of
In Sections 8.2 and 8.3 X,
unless they are explicitly defined to be something else.
If, for some countable ordinal α, the forest class
If, for some
This follows directly from the definitions. Let us consider some
□
For all countable ordinals α the following assertion is true. For all
We show this by induction over α. Let us fix some countable ordinal α and assume that the assertion is true for all smaller ordinals. Let us assume that, for some good function
Essentially the same assertion is true if
For all countable ordinals α the following assertion is true. For all
The proof is very similar to (in fact, easier than) the proof of Lemma 8.8. □
For all countable ordinals α, β the following assertions hold true.
If
For all
Let us assume We prove this by induction over α. Let us assume that for some
□
We note that the condition
The next two assertions are about
For all countable ordinals α, β with
Let
We will not make any use of the following strengthening of the second assertion of Proposition 8.3. We have included it because it is interesting in its own right.
Let X be a second countable topological space, and let
Let For the sake of a contradiction, let us assume that this is false. Then, for all n, there exists a countable ordinal We claim:
Let α be a countable ordinal.
Let
If
The same assertions are true if
We show this by induction over α. Let us fix some countable ordinal α and assume that the assertion is true for all smaller ordinals. Let us assume that First case: Second case: Let us assume that The proof of these assertions for
□
For all countable ordinals α the following assertions are true.
If
For all
We prove the first assertion. The second is proved in the same way. First we note that, if
It is possible that for some points x the sequence of tree classes
The following lemma will be crucial in the following section.
Let α be a countable ordinal. If
Let us assume that
In the first of the two subsections in this section we formulate and prove the results that connect the Weihrauch reducibility
We remind the reader that by Proposition 8.3 all considered forest classes and tree classes exist if the preordered sets
Weihrauch reducibility with parameter
Let X,
If
for all
We prove this by induction over α.
For
For any
Let us consider a countable ordinal
Let us consider some
The following lemma says that if a function f defined on a subset of the Baire space is “locally”
Let
Let us assume that for every
Let
We are going to show this by induction over α. Let us assume that α is a countable ordinal such that Let us prove this claim. Note that Let For each The set V is the disjoint union of the clopen (in the relative topology of We claim that these functions A and B show
Let
This follows directly from Proposition 9.1 and Proposition 9.3. □
The following proposition is a version of Proposition 9.1 for Wadge, strong Weihrauch, and Weihrauch reducibility.
Let X,
Let
if
for all
Let
if
for all
The proof of these assertions is very similar to (in fact easier than) the proof of Proposition 9.1. It is left to the reader. According to Lemmas 5.13.3 and 7.24 the assertion for The case Finally, the case
□
The following proposition is a version of Proposition 9.3 for Wadge, strong Weihrauch, and Weihrauch reducibility.
Let
Let
Let
The proof of this assertion is very similar to (in fact easier than) the proof of Proposition 9.3. It is also straightforward to formulate and prove a version of Lemma 9.2 for According to Lemmas 5.13.3 and 7.11 the assertion for The case Finally, the case
□
If X is a second-countable topological space, Either or there is a (uniquely determined) countable ordinal α such that
Let
Let
Let
This follows directly from Proposition 9.5 and Proposition 9.6. □
Let
In Proposition 10.5 we shall formulate a sufficient condition ensuring that the second of these two cases is true.
In this section we consider functions defined on the Baire space
Functions, forests and their ranks
First we give an upper estimate of the ranks of the forest classes and tree classes associated with a function. Then we show that for every forest class
Let X be a second countable topological space, let
If
for all
We prove both assertions by induction over α.
Let us first consider the case
We come to the case
For the second assertion, let us consider some
We remind the reader once again that all forest classes
Let X be a second countable topological space, let
if
f is
For all
For all
Note that any function We prove this by induction over The set Now, let First case: There does not exist a Second case: There exists a If Let us assume that
Let X be a second countable topological space, let
For every
There exists a countable ordinal α such that for all
The direction
For the proof of the central result in this subsection, Proposition 10.5, we need the notion of a meager set, for which the reader is referred to [11]. The following result is due to Baire.
Let X be a metrizable topological space, and let Y be a second-countable topological space. Let
For the proof see [11, (24.14)], where the assertion is stated under the additional assumption that Y is metrizable. But this assumption is not needed.
Let
It is clear that the condition “for all With any function For the sake of a contradiction, let us assume that the assertion is false. Then there exists a function Claim 1: For all Proof: Let us consider some Claim 2: The set A is a closed subset of X. Proof: Let us assume that A is not closed. Then there exist a point As A is closed, for all Claim 3: The function Proof: For the sake of a contradiction let us assume that there exists some point We have shown that the set
For a preordered set
Let
the
the
Of course, in “ According to Proposition 10.5 for every function Let We give an example of a function
on the one hand and the partial order
In this section we apply the results from Section 9 to relations
Forests and trees and admissible representations
First we show that equivalent representations lead to identical associated forest classes.
Let X be a set with two equivalent representations
That ρ and σ are equivalent means
The following proposition shows that, in the case of a second countable
Let X be a second countable
Let For all open for all For
Remember that for a relation
Let X,
If X,
Remember that by Lemma 11.1 the forest classes
Let us note again that, because Y is finite, the partially ordered set
In the following corollary we consider the special case when the relations are actually functions, and we are content with considering second-countable Let X,
There exists a monotone function
As the proof of Theorem 11.3, but with Lemma 5.2 and Lemma 4.2 instead of Corollary 5.4. □
First we consider relations with range in a finite
Let X,
There exists a relatively continuous relation
Remember that by Lemma 11.1 the forest classes
According to Corollary 5.8 the first condition,
Next we consider the more special case when the range is finite and discrete.
Let X,
There exists a monotone function
“
“
In the following corollary we consider the special case when the relations are actually functions, and we are content with considering second-countable Let X,
There exists a monotone function
First, one should apply Proposition 11.2. Next, one observes that for a function f and some
Again, in the following, on
Let X,
There exists a monotone function
If X,
Remember that by Lemma 11.1 the forest classes
First we note that, because Y is finite, the partially ordered set
“
“
In the following corollary we consider the special case when the relations are actually functions, and we are content with considering second-countable Let X,
There exists a monotone function
The proof is very similar to the proof of Corollary 11.7. One applies Proposition 11.2, Theorem 11.8, and Remark 7.22. □
This example is a continuation of Examples 4.12 and 8.4. In Example 4.12 we defined a continuous function I such that
We still wish to show
This example is a continuation of Examples 4.13 and 8.5. We still wish to show
But we can also prove the claim by computing the forests describing the discontinuity of
We would like to conclude this paper with some open problems.
Let
While this would certainly be interesting, it would still mainly cover only functions with range in a bqo set (and some exceptional other functions) and only relations with a rather restricted range, e.g., with a finite discrete range. Furthermore, for applications the relations
There are a lot of operators on multi-valued functions that are important in the theory of Weihrauch degrees; see [2]. It might be interesting to analyze what effect they have on the tree classes and forest classes associated with multi-valued functions when they are applied to multi-valued functions.
One may of course try to compute the tree classes and forest classes associated with multi-valued functions appearing in the practice of computing. They provide a kind of combinatorial description of the discontinuities of computation problems. It would be interesting to see whether this description turns out to be useful for understanding better the discontinuities of computation problems and for implementing algorithms for them.
