6
$\begingroup$

I have some questions about Turing machines. Is there an embedding method where you embed Turing machines, finite automata into continuous space or graphs? Or are there geometrical approaches to analyze Turing machines or automata? ( embed them into some spaces and apply geometrical tools) Thank you

$\endgroup$
2
  • $\begingroup$ Your question seems a bit vague. Did you google? $\endgroup$ Commented Oct 21, 2018 at 8:25
  • $\begingroup$ There are countably many turing machines. At least I can say that from the point of view of real and comlex analysis, if the usual topologies are used there then there isnt much to talk about. Sadly, a lot of the nicer geometric results rely on real/complex analysis. $\endgroup$ Commented Oct 21, 2018 at 9:05

1 Answer 1

3
$\begingroup$

Let $\mathcal{M}$ be a class of binary functions acting on strings in $\Sigma^*$, along with a "size" function $|\cdot|:\mathcal{M}\to\{1,2,\ldots\}$ with the property that there are only finitely many $M\in\mathcal{M}$ of a given size. Turing machines and finite automata are of this form, where size can be taken as the number of states. Then you can define the following symmetric positive definite kernel on $\Sigma^*$: $$ K_n(x,x') = \sum_{M\in\mathcal{M}, |M|\le n} M(x)M(x'). $$ In words, this counts the number of $M$ on at most $n$ states which accept both $x$ and $x'$. This kernel then embeds $\Sigma^*$ into a corresponding Reproducing Kernel Hilbert Space. One can ask what languages can be defined by linear separators under this kernel. This topic was explored in https://www.sciencedirect.com/science/article/pii/S0304397508004581 and http://www.jmlr.org/papers/volume10/kontorovich09a/kontorovich09a.pdf .

$\endgroup$
1
  • 1
    $\begingroup$ Read only few pages but it seems like a good place to start. I was actually looking for a way to encode my automata/Turing machine to feed them into DNN or other classifier, seems your answer is very relevant, thank you! $\endgroup$ Commented Oct 21, 2018 at 11:53

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.