Peter Steinbach

Python3 fixes

2

I installed the module through `pip3` - which didn't throw an error while doing so. Only after the fact, I found out that this module is python2 only. I started...

lower entry-barrier for beginners

2

When running this cookiecutter, it's quite a challenge to oversee the implications of e.g. all the coverage tools and how to start using this cookie cutter. If there is support...

extracting bruteforce NN class with proper citation

4

Hi @erikbern et al, thanks so much for providing this wonderful venue to see benchmarks of ANN implementations. I am wondering if you would support me in extracting the [BruteForceBLAS](https://github.com/erikbern/ann-benchmarks/blob/a851a311ec8ac2488383fdfb06a185d8bc3e8062/ann_benchmarks/algorithms/bruteforce.py#L33)...

just to prevent misunderstandings

with pytorch `1.2` a dataset of size `0` produces an error. this PR adds a remark on this feature and how a user can resolve this.

this may help understand readers why the code converts from numpy to pytorch and back again

I hope this helps understanding better where these values come from.

Hi, thanks for providing this wonderful package. I took the liberty and added a flag to the CLI of `glow.py` in order to download e.g. the MNIST dataset. I hope...

using exact similarity search

1

Thank you for providing the coding and implementation for density and coverage. It is awesome to have the code ready for use in practice. I cite the paper whenever possible....

a simple change, that splits out the density and coverage function from `compute_prdc`. This helps users to omit the CPU cycles required to evaluate precision and recall.