2 ms·
Has there been any progress on https://www.kaggle.com/c/integer-sequence-learning https://www.kaggle.com/c/integer-sequence-learning?
by uyt 5y ago
Has there been any progress on https://www.kaggle.com/c/integer-sequence-learning https://www.kaggle.com/c/integer-sequence-learning?
- lifthrasiir 5y agoThere is also an attempt [1] to collect and (largely) synthesize a program for every OEIS sequence in an intentionally constrained DSL. [1] https://github.com/ckrause/loda https://github.com/ckrause/loda
- forgotpwd16 5y ago>In total, there are currently more than 26,000 programs available Wow, lot of work has gone to it. Why the decision to utilize an assembly language?
- lifthrasiir 5y agoAs far as I can tell its assembly language is designed for fast synthesis and check (e.g. all programs are known to terminate). There is a "miner" generating programs and checking against existing sequences 24/7, a Twitter bot is also available [1]. [1] https://twitter.com/lodaminer https://twitter.com/lodaminer
- glangdale 5y agoThanks for this link. I am working on a superoptimizer and once had the thought that OEIS would be a cool place to look for workloads. It's gratifying that someone else is doing this and having some success.