Austin Tripp

I found that in your example code, you loaded the model, then ran the variable initialisation, which effectively erased the loaded model. I changed the code to load the model...

One of the cards had the typo `uniqe` instead of `uniq`. The card ID was 1484731420055. I fixed this typo and added the collection again to the apkg file.

From your `README.md` file it was not 100% clear exactly what command(s) I should run if I want to get results for a new baseline algorithm and extract the AUC...

In `run.py`, some information is printed to the terminal, but not the AUC-Top10 (i.e. the main metric which is suggested by your paper). This PR harmonizes the info outputted to...

Add the "minmax" kernel

3

#### What does this implement/fix? Explain your changes This PR provides a memory-efficient implementation for the minmax kernel from my personal BO repo. It uses a formulation from Ioffe (2010)...

Inconsistent / unclear python version requirements?

1

Currently the python versioning of this project is unclear: the README has no info, the supplied conda env says python 3.7, while the internal `setup.py` says `python>=3.8`. Maybe this should...

After playing around with this code as a starting point for my own GA algorithm, I think that the reproduce function of the graph GA modifies the parent mol objects,...

Raise error if "initialize" method is called in eval mode.

9

This PR is in response to #1556, wherein updating kernel hyperparameters while in eval mode leads to incorrect predictions because the cached values in `model.prediction_strategy` are not updated. As suggested...

CI could run tests, but at the very least should check linting.

Indexing error makes many fingerprint kernel implementations incorrect

3

In a variety of different fingerprint kernels, `x1` and `x2` are compared with something like ``` x1[-1] x2[-1] ``` which only compares the last element in the kernel matrix, and...