Skip to content
This repository has been archived by the owner on Jan 5, 2024. It is now read-only.

What do the two aggregated score numbers mean? #81

Open
justinborromeo opened this issue Apr 21, 2021 · 1 comment
Open

What do the two aggregated score numbers mean? #81

justinborromeo opened this issue Apr 21, 2021 · 1 comment

Comments

@justinborromeo
Copy link

image

Also, what are raw and ceiling?

@mschrimpf
Copy link
Member

mschrimpf commented Apr 21, 2021

The two numbers correspond to the aggregate score and an estimate of the error of that score. More specifically they correspond to the two coordinates center and error in the aggregation dimension. For instance, for the standard neural predictivity metric, the center corresponds to the mean over cross-validation splits and median over neuroids. These numbers are further normalized by the ceiling (see below).

ceiling is the reliability of the benchmark; since the data come from noisy biological measurements, this tries to estimate the maximum value we would expect a model to score (in this case ~0.82).

raw in this case is the aggregate correlation (mean over splits, median over neuroids) before normalizing with the ceiling. See here for how this particular benchmark normalizes scores by the ceiling. In effect, this is just another Score object and itself contains another raw object which should be the per-split, per-neuroid correlation values, i.e. before aggregating them into a single number.

More details can be found in the benchmarks and metrics examples.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants