[seqfan] Re: Using AI to help Superseeker guess formulas for sequences

[Antti Karttunen]

Just a thought (haven't yet but just skimmed that article, but this
came to my mind immediately): What if, instead of limiting the search
space to a fixed set of the most common mathematical operators (+, x,
sin, etc) in the constructed unary/binary-trees, one could construct
similar expressions from any OEIS sequences, tables, and their
transformations?  As we know, one-dimensional sequences can be viewed
as N -> Z functions (unary operators), and square arrays as N x N -> Z
functions (dyadic operators).
This is one of the reasons I am for long b-files (until we get
reliable and fast on-the-fly generation of the terms for most
sequences, that is).

Best regards,

Antti


On 10/4/20, Neil Sloane <njasloane at gmail.com> wrote:
> Marc, Thank you for this reference, which is extremely relevant to my
> question!
>
> Lample and Charton, (from FaceBook AI Research)
> "Deep Learning for Symbolic Mathematics"
> https://arxiv.org/pdf/1912.01412.pdf
>
>
> After I've digested it, perhaps I will talk to them about sequences!
>
> Best regards
> Neil
>
> Neil J. A. Sloane, President, OEIS Foundation.
> 11 South Adelaide Avenue, Highland Park, NJ 08904, USA.
> Also Visiting Scientist, Math. Dept., Rutgers University, Piscataway, NJ.
> Phone: 732 828 6098; home page: http://NeilSloane.com
> Email: njasloane at gmail.com
>
>
>
> On Sun, Oct 4, 2020 at 1:08 PM Marc LeBrun <mlb at well.com> wrote:
>
>> >=Neil Sloane
>> Why not use AI to strengthen Superseeker?, I keep saying
>>
>>
>> Indeed.  Want to mention this again, for anyone who might pursue this
>> further:
>>
>> "Deep Learning for Symbolic Mathematics"
>> https://arxiv.org/pdf/1912.01412.pdf
>>
>> Not necessarily the answer, but perhaps it may inspire something
>> interesting?
>>
>>
>> --
>> Seqfan Mailing list - http://list.seqfan.eu/
>>
>
> --
> Seqfan Mailing list - http://list.seqfan.eu/
>