More details on how to set the algorithm's hyperparameters in the configuration file ?

[hugovaysset]

Hello,

First of all, thank you for your work on this very useful library.

I would like to have more details about the hyperparameters of the algorithm (especially of the Hypothesis model) and how I need to set them in the config file regarding my specific data. I work on cell tracking, and typically have several hundreds of images and around a hundred cells within each image. For now, I've only used the algorithm with the default json config file given in the tutorial and this works fine as long as the number of images remains fairly low (~100). When I go a bit beyond (~300 images), the "Optimizing" step never ends. This is must be due to a poor optimizer configuration, as underlined in Issue #13.

I understand that I should change the Hypothesis Model hyperparameters in the config file, but the problem is that I couldn't find any detailed information about what is each hyperparameter useful for, and how it should be regarding the data (not enough details in the wiki or in the referenced papers). Did I miss something, or could it be possible to have more information regarding this ?

Thank you!

[quantumjot]

Hi @hugovaysset - thanks - I hope it's proving useful!

We've been meaning to update the documentation for a while. There's a new CONFIGURATION.md in the root directory with some details now. Can you let me know which parameters in particular you're having trouble with?

[hugovaysset]

Hi @quantumjot, thanks for the quick answer!

Actually, my questions are very broad and basic about the algorithm that is used in btrack.

I solved the computation time issue by just removing the optimisation step. What is the point of this step ? In other words, what do I loose when I don't do it ? I suppose that this step can substantially increase the quality of the tracking results, so I would like to better understand it. In issue #13, you said "you probably need to change some parameters in the config to better to describe what you are trying to track and what you care about.". Could you please tell me more about which parameters are the most important ones ? Are they more in the Motion model or in the Hypothesis model? I suppose that the number and the type of hypotheses that we test must be important, should I tune them to save computational time? Is the precise value of each numerical parameter in the Hypothesis model (lambda_time, lambda_dist, eta, etc.) really important?

Additionally, I would also be interested in knowing what could be done with the Object model (that is empty in the config file given in the documentation). Is it possible to describe the cells that we want to track, in order to integrate not only the position but also for example the surface of the cells to improve the quality of the tracking results ?

Thanks again.