m Tag: Visual edit |
m (correction) |
||

Line 23: | Line 23: | ||

# Every tree at position ''k'' (for all ''k'') has no more than ''k'' + ''n'' vertices. |
# Every tree at position ''k'' (for all ''k'') has no more than ''k'' + ''n'' vertices. |
||

# No tree is homeomorphically embeddable into any tree following it in the sequence. |
# No tree is homeomorphically embeddable into any tree following it in the sequence. |
||

− | Adam P. Goucher |
+ | Adam P. Goucher stated the following properties of this function without proofs:<ref>Adam P. Goucher, [https://cp4space.wordpress.com/2012/12/19/fast-growing-2/ TREE(3) and impartial games]</ref> |

# tree(n) has a "growth rate" comparable to that of [[SVO|\(f_{\vartheta(\Omega^\omega)}(n)\)]] in the fast-growing hierarchy. |
# tree(n) has a "growth rate" comparable to that of [[SVO|\(f_{\vartheta(\Omega^\omega)}(n)\)]] in the fast-growing hierarchy. |
||

# TREE[3] > tree<sup>tree<sup>tree<sup>tree<sup>tree<sup>8</sup>(7)</sup>(7)</sup>(7)</sup>(7)</sup>(7) |
# TREE[3] > tree<sup>tree<sup>tree<sup>tree<sup>tree<sup>8</sup>(7)</sup>(7)</sup>(7)</sup>(7)</sup>(7) |
||

− | A larger lower bound has since been found. Define \(\text{tree}_2(n)=\text{tree}^n(n)\) and \(\text{tree}_3(n)=\text{tree}_2^n(n)\). Then \(\text{TREE}[3] > \text{tree}_3(\text{tree}_2(\text{tree}(8)))\). The latter expression is equal to tree<sup>tree<sup>tree<sup>...<sup>tree(n)</sup>...</sup>(n)</sup>(n)</sup>(n) with n layers, where \(n = \text{tree}^{\text{tree}(8)}(\text{tree}(8))\).<ref>[http://math.stackexchange.com/questions/313134/how-does-tree3-grow-to-get-so-big-need-laymen-explanation How does TREE(3) grow to get so big? Need laymen explanation]</ref> |
+ | Since he stated them as if they were obvious facts, they were believed to be verified by himself in this community for a while. A larger lower bound has since been found. Define \(\text{tree}_2(n)=\text{tree}^n(n)\) and \(\text{tree}_3(n)=\text{tree}_2^n(n)\). Then \(\text{TREE}[3] > \text{tree}_3(\text{tree}_2(\text{tree}(8)))\). The latter expression is equal to tree<sup>tree<sup>tree<sup>...<sup>tree(n)</sup>...</sup>(n)</sup>(n)</sup>(n) with n layers, where \(n = \text{tree}^{\text{tree}(8)}(\text{tree}(8))\).<ref>[http://math.stackexchange.com/questions/313134/how-does-tree3-grow-to-get-so-big-need-laymen-explanation How does TREE(3) grow to get so big? Need laymen explanation]</ref> |

=== Values for \(\text{tree}(n)\) === |
=== Values for \(\text{tree}(n)\) === |

## Revision as of 03:43, 7 May 2020

*Not to be confused with Exploding Tree Function, or Friedman's finite trees.*

The

**TREE sequence**is a fast-growing function arising out of graph theory, devised by mathematical logician Harvey Friedman.

^{[1]}

^{[2]}Friedman proved that the function eventually dominates all recursive functions provably total in the system \(\text{ACA}_0\)+\(\Pi_2^1\)-\(\text{BI}\).

^{[3]}

The first significantly large member of the sequence is the famous **TREE[3] **(sometimes written as TREE(3)), notable because it is a number that appears in serious mathematics that is larger than Graham's number.

## Contents

## Definition

Suppose we have a sequence of *k*-labeled trees T_{1}, T_{2} ... with the following properties:

- Each tree T
_{i}has at most*i*vertices. - No tree is inf preserving and label preserving embeddable into any tree following it in the sequence.

Kruskal's tree theorem states that such a sequence cannot be infinite. Harvey Friedman expanded on this by asking the question: given some *k*, what is the maximum length of such a sequence?

This maximal length is a function of *k*, and is dubbed TREE[*k*]. The first two values are TREE[1] = 1 and TREE[2] = 3. The next value, **TREE[3]**, is famously very large. It vastly exceeds Graham's number and n^{n(5)}(5)^{[4]}. Chris Bird claimed that \(\text{TREE}[3] > \{3, 6, 3 [1 [1 \neg 1,2] 2] 2\}\), using his array notation.^{[5]}

TREE[n] is strongly believed to grow at least as fast as \(f_{\vartheta(\Omega^\omega\omega)}(n)\) in the fast-growing hierarchy, making it quite sizable even to a googologist. If this belief is indeed correct, then the TREE function is more powerful than Kirby-Paris hydras and Goodstein sequences, but weaker than subcubic graph numbers and Buchholz hydras. For a look into one of the issues of estimating high growth rates, check this blog post by Jason.

## Weak tree function

Define \(\text{tree}(n)\), the **weak tree function**, as the length of the longest sequence of 1-labelled trees such that:

- Every tree at position
*k*(for all*k*) has no more than*k*+*n*vertices. - No tree is homeomorphically embeddable into any tree following it in the sequence.

Adam P. Goucher stated the following properties of this function without proofs:^{[6]}

- tree(n) has a "growth rate" comparable to that of \(f_{\vartheta(\Omega^\omega)}(n)\) in the fast-growing hierarchy.
- TREE[3] > tree
^{treetreetreetree8(7)(7)(7)(7)}(7)

Since he stated them as if they were obvious facts, they were believed to be verified by himself in this community for a while. A larger lower bound has since been found. Define \(\text{tree}_2(n)=\text{tree}^n(n)\) and \(\text{tree}_3(n)=\text{tree}_2^n(n)\). Then \(\text{TREE}[3] > \text{tree}_3(\text{tree}_2(\text{tree}(8)))\). The latter expression is equal to tree^{treetree...tree(n)...(n)(n)}(n) with n layers, where \(n = \text{tree}^{\text{tree}(8)}(\text{tree}(8))\).^{[7]}

### Values for \(\text{tree}(n)\)

It can be shown that \(\text{tree}(1) = 2, \text{tree}(2) = 5\) and \(\text{tree}(3) \geq 844424930131960\). \(\text{tree}(1)\) uses the sequence:

(()) ()

\(\text{tree}(2)\) is a bit larger, we have two longest sequences that go as follows:

((())) (()()()) (()()) (()) ()

Otherwise:

(()()) (((()))) ((())) (()) ()

Determining the exact value for \(\text{tree}(3)\) is much harder, since there are a lot of sequences to check, and each of these is very long.

Friedman has defined an FFF(k) function, which is equal to tree(k+1).^{[8]}

## Alternative notations

*(This alternative has yet to be formally verified.)*

Trees are tricky to visualize without drawing them out, so we shall devise a more compact way of representing them. Consider a language which has various kinds of parentheses such as `()`

, `[]`

, `{}`

etc. Parentheses always come in pairs and can nest within each other. Within a larger node, they may be concatenated. For example, the following strings are valid in this language:

[] ([]) {[()]()} [(){[[]]}(){(())[]}]

Suppose we have a string *A*. We shall call a pair of corresponding parentheses a **node**, in deference to the original tree construction. Define a **child** of a node to be a node that is nested one level deep within the original node. For example, take the string `{[()()][][()]}`

; the children of the node represented by `{}`

are the nodes represented by `[]`

, but not the nodes represented by `()`

.

Call a node **deletable** if it has fewer than two children. For example, in the string `{[()()][][()]}`

, the `()`

nodes are all deletable, as are the latter two `[]`

nodes, but not the first `[]`

or the `{}`

. In the string `([(()())])`

, the `[]`

node is deletable.

We say a string A is **reducible** to a string B if A can be transformed into B by removing deletable nodes. A string A is reducible to a string B if and only if the tree represented by B is homeomorphically embeddable in the tree represented by A.

With all this in mind, we can create a function which should correspond with the original definition of TREE[k] assuming this notation was derived correctly. Suppose we have a sequence of strings with the following properties:

- You may only use
*k*types of brackets. - The first string has at most one pair of brackets, the second string has at most two pairs of brackets, the third string has at most three pairs of brackets, etc.
- No string is reducible to an earlier string.

TREE[k] is the maximal length of the sequence.

For *k* = 1, the optimal sequence has only one member: `()`

.

For *k* = 2, the optimal sequence has only three members: `()`

, then `[[]]`

, then `[]`

.

### Weak tree function

Suppose we have a sequence of strings with the following properties:

- You may only use
`()`

and no other types of brackets. - The first string has at most 1 +
*k*pairs of brackets, the second string has at most 2 +*k*pairs of brackets, the third string has at most 3 +*k*pairs of brackets, etc. - No string is carvable from a later string.

tree(*k*) (the weak tree function) is the maximal length of the sequence.

## Videos

## Sources

- ↑ H. Friedman, [FOM] 273:Sigma01/optimal/size
- ↑ H. Friedman, [FOM] n(3) < Graham's number < n(4) < TREE[3]
- ↑ http://www.cs.nyu.edu/pipermail/fom/2006-March/010279.html
- ↑ How large is TREE(3)
- ↑ Chris Bird, Beyond Bird's Nested Arrays II, page 10
- ↑ Adam P. Goucher, TREE(3) and impartial games
- ↑ How does TREE(3) grow to get so big? Need laymen explanation
- ↑ http://www.cs.nyu.edu/pipermail/fom/2006-June/010627.html

## See also

**By Harvey Friedman:** Mythical tree problem · Friedman's vector reduction problem · Friedman's finite ordered tree problem · block subsequence theorem n(4) · Friedman's circle theorem · **TREE sequence** TREE(3) · subcubic graph number SCG(13) · transcendental integer · finite promise games · Friedman's finite trees · Greedy clique sequence**Miscellaneous:** Factorial · Folkman's number · Exploding Tree Function · Graham's number · fusible number