Alejandro Pulver

How about building an index for read-only random access? You can store compressed block sizes/positions in a tree so you can know where any position starts in log(n) operations. The...

@overheadhunter I agree that what I proposed is more of a work around / hack and not a proper solution. To allow for random access the vfs would have to...

> So access to data would be one query to your ledger and then one query for data right? That seems like random access to me. You have O(1) runtime....

This seems to happen with any matrix equivalent to `numpy.diag(a,b,b,a)` when a and b are lower than 20. I believe `svdLAS2SA` could be returning something different in these cases. What...

Could you please confirm that `svdResult->d` is 2, as it should be? Edit: don't worry, I don't think that's the case. I'm getting d=0 and S=[0,0] from SVDLIBC. Either I'm...

When enabling SVDVerbose in svdlib.c, I get the following output when it works: ``` In [2]: sparsesvd.sparsesvd(scipy.sparse.csc_matrix(np.diag([1,15,15.1,1])), 2) SOLVING THE [A^TA] EIGENPROBLEM NO. OF ROWS = 4 NO. OF COLUMNS...

Sorry for the delay. I don't know if it's related, but at least 0.1.8 behaves the same in my machine. The only difference is that it doesn't throw an exception,...

Thank you for spending so much time on this issue. I'm not really into numerical methods, so I may not be able to help. But maybe there is a paper...

I don't think I can debug it. Maybe an exception could be raised in those cases, or adding a tiny random noise to make it work (at least close to...

me too, with latest JRuby 1.7.3 and drip from git