Nice to see the EteRNA bots activated
Recently I received messages in my inbox about Viennabot not being able to solve some of my puzzles (I have updated the puzzle list in my profile accordingly).
I was wondering, did you only activate Viennabot or did you activate the other two as well?
Did you activate it (or them) on all puzzles, or did you run a certain section of puzzles through the bot?
If the other bots were not employed this time, when will you test those?
In order to interpret the bot results it would be nice with some info regarding these questions.
By the way, the bot status states inactive although the bot is running?
I’m really looking forward to the results of the bot activation.
It would be interesting to have a list of solved/unsolved puzzles + extra info:
1 number of bases in the design, ratio between basepairs and free bases
2 number and nature of constraints
3 differences between bot algorithms (how they are designed) and difference in solved/unsolved between the algorithms
E.g. Why could the ViennaRNA bot solve Seagull (57 bases) but not Crossroads (32 bases)? Are the lesser degree of constraints in Seagull a key factor or has topologigal complexity something to do with it?
I’d be really interested in the answers to some of these questions too
Especially: where do the bots come from (who designed them)? Are they similar in their approaches to the problem or do they approach them differently, and how?
And it would be great to have a list of puzzles that each one did/didn’t solve, rather than having to check the comments on each individual puzzle. Maybe at least their profile pages could show a list of solved puzzles like other player profile pages?
From what I understand, the bots aren’t under the same GC/GU/UA constraints that players are (I think they aren’t designed to work that way) so they’re just solving the basic shapes. And I have seen some comments from InfoRNA on the older player puzzles, so I think it’s active too and working its way through player puzzles a little behind ViennaRNA. It would be really interesting to be able to compare which puzzles each one can solve
If they only solve shapes it is truly mind-boggling that the ViennaRNA bot was unable to solve Crossroads!
Bots are not “officially” in yet. We are still going over testing phase. They’ll become active once we are done with the tests.
Vienna RNA bot basically uses stochastic search, which means it’s behavior is largely random. It could solve very hard puzzles easily with luck, or it could get stuck in very easy puzzles.
Other two algorithms, InfoRNA and RNASSD were much stronger than Vienna RNA in our tests. So these are two main challenges you can expect.
Ding - you are right. Bots aren’t under same pair constraints like people. These bot algorithms are fundamentally not designed for such constraints - all they can do is to just solve for the shape (maybe with some bases locked, but not under constraints like players).