[to_review] parameter for 2d data generation with 'regression' label

[BuenoRuben]

it allows to choose mean and covariance matrix for the generated data

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Comparison is base (2f2bb85) 89.08% compared to head (6bf1a86) 89.07%.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #79 +/- ## ========================================== - Coverage 89.08% 89.07% -0.01% ========================================== Files 41 41 Lines 2731 2738 +7 ========================================== + Hits 2433 2439 +6 - Misses 298 299 +1 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[rflamary]

While you are at it @BuenoRuben could you please scale the output y so that it has reasonable values? it is very small at the moment so that when i computed MSE I had something around 1e-6

It should have a variance around 1 if possible (but do not scale it for each seed it would introduce some other shift)

[BuenoRuben]

i am doing it right now

[BuenoRuben]

While you are at it @BuenoRuben could you please scale the output y so that it has reasonable values? it is very small at the moment so that when i computed MSE I had something around 1e-6

It should have a variance around 1 if possible (but do not scale it for each seed it would introduce some other shift)

it should be fine now

[tgnassou]

I don't know what happened, but your examples show that label distributions are changing between source and target. Look at Rémi's DA course. There is an example of a regression dataset for covariate shift etc. Maybe it can help you to see what it should look like.

[BuenoRuben]

that's corrected