search

NVlabs / FPSci

Aim Training Experiments
Other
70 stars 23 forks source link

Staircase Stimulus #390

Open bboudaoud-nv opened 2 years ago

bboudaoud-nv commented 2 years ago

This issue is a placeholder for discussion of staircase stimulus in FPSci.

Issues to address include:

  1. How to specify the parameter, step size, and maximum step count for the staircase?
  2. How to limit invalid configuration caused by experimenter misunderstanding of parameters (i.e. min step < controllable step, parameter cannot be staircased)?
  3. How to incorporate into existing FPSci configuration (different session type)?
  4. How to manage bugs that might arise in the field but not be present in pilots (long run time/invalid results)?
  5. How to implement/where we can borrow a valid staircase stimulus generator algorithm from?

This is not currently planned for development in FPSci, but is worth discussing. We would likely plan to have an early implementation complete around October 2022.

bboudaoud-nv commented 2 years ago

The motivation for this addition is staircasing frameRate and frameDelay for frame rate/latency JND experiments, but its worth considering where else this could be used.

Generally speaking if it is worth developing this feature, it should generalize well. FPSci does not typically include features/code to support 1-off features.

bboudaoud-nv commented 2 years ago

As part of this development I think we will need to add a (set of) FPSci ConfigParameter classes that generalize our configuration and provide some meta-level information about each configuration parameter. This will be a fairly major facelift to the configuration system that will occur in its own branch prior to implementation of the staircase-style session.

bboudaoud-nv commented 2 years ago

Alternatively to a ConfigParameter we can potentially use the Any config structure to implement this feature. This is closer to the idea of having per trial configuration, which we could unite with this approach.