Issa Rice
Issa Rice
More progress: see 5136c4b21bbf1aad54cef0cfcc581b286ef39c12. By building up the `donor_donee_pairs` and `donor_donee_pairs_2` tables incrementally, it looks like the first few calls can be made really fast. So if we only care...
Here is progress so far:  The plotting script is [here](https://gist.github.com/riceissa/4506fa16d4dbae794387979843e74c22). Run it like `./plot.py data` where `data` is the output of the the calls to `updateSimilarityDonee`. So it looks...
More progress:  The script actually crashed because one of the donees was too long (I didn't realize the `donee` column was `varchar(200)` while `donor` is `varchar(100)`). It seems pretty...
Some things to do: - [x] Review previous work in this issue and elsewhere to refresh my memory of what I was even doing - [ ] @vipulnaik suggested breaking...
I've now implemented the similarity computation in Python for donors at https://github.com/vipulnaik/donations/blob/e2b66172d8b6d66bbb797b25dedde32e95b39780/similarity/similarity.py and for donees at https://github.com/vipulnaik/donations/blob/e2b66172d8b6d66bbb797b25dedde32e95b39780/similarity/similarity_donee.py (for donees I loop over all donees because I was testing how long...
Ok here is an idea: if we consider a single donee like MIRI, which other donees will have non-zero similarity with MIRI? The only way to have non-zero similarity is...
Another idea (which can work in combination with the previous one) is to start the similarity computation for a donee when its page is first requested, storing the result in...
It seems like most donees have several thousand other donees that have non-zero similarity (this is mostly because large foundations+donees of such foundations dominate the donations data), so that's several...
According to Fixer.io, the conversion rates on 2019-05-31 (date of the blog post) are: * 1.119845 USD = 1 EUR * 1.516663 CAD = 1 EUR So 1 CAD =...
At I see: > Our new grants database filters are still being finalized. Until then, you may see a reduced number of results. Using a keyword search might help you...