Definition
Given a function, f:X→Y and another function, w:X→W (I have chosen W to mean "whatever") we can say:
- f may be factored through w
if f and w are such that:
- ∀x,y∈X[w(x)=w(y)⟹f(x)=f(y)]
- (this is the same as: ∀x,y∈X[f(x)≠f(y)⟹w(x)≠w(y)])
Then f induces a function, ˜f such that f=˜f∘w, or more simply that the following diagram commutes:
Note:
- ˜f may be explicitly written as ˜f:W→Y by ˜f:v↦f(w−1(v))
- Or indeed ˜f:=f∘w−1
- This is actually an abuse of notation as w−1(x∈W) is a subset of X, however it is safe to use it because (as is proved below) f of any element of w−1(x∈W) for a given x is the same.
- The function ˜f is unique if w is surjective
Points to remember
- Remembering the requirements:
- We want to induce a function ˜f:W→Y - if w(x)=w(y) then ˜f(w(x))=˜f(w(y)) just by composition.
- If f(x)≠f(y) we're screwed in this case. So it is easy to see that we must have [w(x)=w(y)]⟹[f(x)=f(y)] otherwise we cannot proceed.
Proof of claims
[Expand]
Claim: the induced function, ˜f exists and is given unambiguously by ˜f:v↦f(w−1(v))
Existence
- Let ˜f:W→Y be given by: f:v↦f(w−1(v)) - I need to prove this is a Function
- This means I must check it is well defined, a function must associate each point in its domain with exactly 1 element of its codomain
- Let v∈W be given
- Let a∈w−1(v) be given
- Let b∈w−1(v) be given
- We know ∀a∈w−1(v) that w(a)=v by definition of w−1
- This means w(a)=w(b)
- But by hypothesis w(a)=w(b)⟹f(a)=f(b)
- So f(a)=f(b)
- Thus given an a∈w−1(v), ∀b∈w−1[f(a)=f(b)]
- We now know (formally) that: (given a v) ∃y∈Y∀a∈w−1(v)[f(a)=y] - notice the ∃y comes first. We can uniquely define f(w−1(v))
- Since v∈W was arbitrary we know ∀v∈W∃y∈Y∀a∈w−1(v)[f(a)=y]
- We have now shown that ˜f can be well defined (as the function that maps a v∈W to a y∈Y.
- To calculate ˜f(v) we may choose any a∈w−1(v) and define ˜f(v)=f(a) - we know f(a) is the same for whichever a∈w−1(v) we choose.
- So we know the function ˜f:W→Y given by ˜f:x↦f(w−1(x)) exists
This completes the proof[1]
[Expand]
Claim: if w is surjective then the induced ˜f is unique
Uniqueness
- Suppose another function exists, ˜f′:W→Y that isn't the same as ˜f:W→Y
- That means ∃u∈W:[˜f(u)≠˜f′(u)]
- However for both ˜f and ˜f′ we have the property of f=˜f∘w=˜f′∘w so:
- By hypothesis we have: ∀x∈X[f(x)=˜f(w(x))=˜f′(w(x))] however we know:
- ∃x′∈X[w(x′)=u] and ˜f(u)≠˜f′(u), this means:
- f(x′)=˜f(w(x′))≠˜f′(w(x′)) - which contradicts the hypothesis.
- However if w is not surjective, then the parts of the domain on which ˜f and ˜f′ disagree on may never actually come up; that is to say:
- ∀x∈X[˜f(w(x))=˜f′(w(x))] as w:X→W may never take an x∈X to a z∈W where ˜f(z) and ˜f′(z) differ; but they could still be different functions.
This completes the proof[1]
- Notes:
- Notice that if w is not surjective, the point(s) on which ˜f and ˜f′ disagree on may never actually come up, so it is indeed not-unique if w isn't surjective.
References
- ↑ Jump up to: 1.0 1.1 This is my (Alec's) own work