Features per Level vs Features per Entry

[nvlabs]

Features per Level vs Features per Entry

[jc211]

I think I may understand now what is meant. I think N_FEATURES_PER_LEVEL means more like N_FEATURES_PER_ENTRY_PER_LEVEL and n_features is then N_FEATURES_PER_ENTRY once all levels are concatenated. Still a bit confusing to me because when reading it, it sounds like the whole level only has 2 features and I have to keep reminding myself that is not what is meant. I don't have any clear suggestions but perhaps DIM_ for dimension is more appropriate than N_.

If this is too much of a nitpick, please close the issue. In any event, I'm enjoying going through the code. Thanks for publishing it.

[Tom94]

I think you're right about N_FEATURES_PER_ENTRY being the better choice in any case. It's clearer in its meaning + we're already using this terminology in the instant-ngp paper.

The idea of naming it n_features_per_level was that the output dimensionality of the encoding ends up as: n_features_per_level * n_levels.