fkarg
commented
8 years ago we got more ideas on JHSued follow-up
Open fkarg opened 9 years ago
fkarg
commented
8 years ago we got more ideas on JHSued follow-up
We spoke about it earlier, and here's the issue: when the data-sets are getting bigger, we need some improvements, our current setting is efficient only with small datasets, and we don't have a database or anything yet. The general idea was that the frontend was written in python, yes? Well, I'll think about it more and might implement some way to actually do it, as well as using some kind of real database. Another Issue I thought about was: when we someday are going to have big datasets, the haskell part is going to take forever, so I had the idea to make a general A* search only with connections in general from the start to the end with some kind of tolerance, meaning I won't look at the probabilities at all but only at the nodes, and how they are connected in general (if at all). and the part I'd give to the actual algorithm wouldn't be one of the n ways I found, but only the nodes included In the n best ways, meaning the Haskell part wouldn't need to make that big a tree, and could use the probabilities either directly from the db or given with the nodes (currently JSON). Or, you could implement something like this too, I won't stop you ^^
But for the sake of divide and conquer I'm gonna create own isses for these, with not that much explanation each. I request closing this issue with the last one here mentioned: #15 (Database) #16 (pre- A* the dataset, and only give relevant information to the haskell part)