Which hyperparameter(s) should take highest priority?

[sgbaird]

• Maybe use get_sites_in_sphere to calculate pairwise distance matrix, so always 52 sites #11
• Consider having certain feature ranges as tunable parameters #13
• Should we use the cubed root of the volume instead of the volume? #41
• Should we set the lower bound of the xtal2png ranges as 0 or as the minimum across all MP compounds? #42
• Write function for adding sensible perturbations to the array representation #64 (applicable to conditional generation)
• Permutation invariance - site ordering and data augmentation #75 (i.e. lexicographic sorting of sites)
• consider composition matrix in place of top-left zeros #76
• Consider using RGB image instead of grayscale to allow for averaging of atomic numbers #81 (i.e. 3 channels instead of 1)
• consider kwarg for mapping noble gases to a "nearby" element or (better) using different elemental featurizer #82 (allow for using different scales, e.g. Modified Pettifor via element-coder)
• consider reconstructing fractional coordinates from distance matrix (via MDS or neural network) #83
• Consider using Wyckoff positions instead of atomic sites #94

[sgbaird]

@sparks-baird/xtal2png, feel free to vote and/or leave a comment. Since the complexity/bug-susceptibility of a codebase tends to increase exponentially with the number of features that are supported but we also don't know for sure which hyperparameters in advance are going to be most useful, I'd like some help prioritizing where the effort goes.