A note on the pre-trained model called "race_no_bias_angle_4"

[gingjehli]

I’ve been working with the pre-trained model called "race_no_bias_angle_4" and noticed that it allows me to both fix the starting point (z) and let it vary. Considering the name of the model includes "no_bias," I wonder if it would make more sense to restrict users from fixing the starting point? Alternatively, one could clarify in the tutorial what "no_bias" specifically refers to in this context? Or maybe even explain what happens if this parameter is not specified at all.

[gingjehli]

also, for a non-biased model: does z have to be fixed at 0.5 or at 0?

[AlexanderFengler]

We should probably find a better naming convention for the race models.
@gingjehli the notion of bias doesn't really apply for these models the same way.

The models marked no_bias allow estimation of this parameter (wouldn't recommend it, because of identifiability issues), the no_z version doesn't have the bias parameter to begin with (trained on having that fixed at 0.5.

Upon reflection, I would probably fix it at 0 instead.

The basic race_.. models will allow you to estimate drift-wise biases, in which case the parameter actually serves the purpose of it's name.