Meeting 13 - GeorgeIniatis/Blood_Brain_Barrier_Drug_Prediction GitHub Wiki

Meeting Minutes

This meeting mainly focused on discussing error analysis

Comments made by supervisor:

There is no clear systematic process to perform error analysis. It is an unspecified task
Confusion matrices for multiclass problems could work very well but not for binary problems
Unsupervised learning on the mistakes could provide useful insights. (PCA os something like K-Means)
The dataset could definetely contain mislabelled drugs/compounds but given the nature of the problem we cannot identify these with any confidence
It would definetely be something nice to have to discuss especially if patterns can be recognised but not a huge amount of time should be spent making it work