Additional Challenge - Cell Migration (Partially Observed System)
We are happy to announce an exciting additional challenge for Cside 2018!
For the purposes of the event, we have currently made the assumption that all the output for Model 3 is observable (global inhibitor, local activator, local inhibitor, stimulus strength, x co-ordinate and y co-ordinate of each finite element node). Whilst this is unlikely to be true in practice, the reduced complexity of the inference task makes it more likely that tackling this model is achievable during the duration of the event.
In order for the application to be closer to the data that is obtainable in practice, we will additionally make data available from Model 3, where only the x and y co-ordinates are observed. This challenge is open to any competitor and will not affect how the results for Model 3 where all output is observed are calculated i.e. data from Model 3 with all output observed AND data from Model 3 with only the x co-ordinate and y co-ordinate will both be available and the results will be ranked independently. The reduced preparation time also more closely mimics an application in the real world.
There is no obligation to submit estimates for this additional challenge, whether you are submitting estimates for Model 3 with all output observed or not. This additional challenge gives us the added benefit of allowing your work to be more similar to real world applications, giving us better insight into the methods’ performance. It also has the advantage of making the follow up publication from the event stronger.
For this additional challenge, only one submission (the person whose method performs the best) will be selected for oral presentation on the conference day (26th November 2018) and will be awarded a £150 Amazon Voucher as a prize.
Please see the section on Model 3 for full details of the model set-up - remembering that this additional challenge will provide data for only the x and y co-ordinates (all other settings will be the same).
Note: The competition data for Model 3 with all output observed, and the data for this additional challenge, will be generated using different coefficients for the parameters of the model.