DLquantification

Repository for researching deep learning approaches for quantification.

GMNet

GMNet implementation is based on pytorch. It is a neural network that focus on the quantification problem. It uses gaussian mixture models to represent the samples.

Experiments

All the experimentation done with GMNet (HistNet or other DL approaches for quantification) lives in the folder ˋexperimentsˋ. You can check the corresponding documentation here

• The experiments related to training with GMNet (or other DL approaches for quantification) are in the folder train_lequa.

• The experiments using validation data with the labels provided by the organizers [1](section "Impact of data availability on quantification methods") are in the folder comp_trad_DL_meth.

  [1] The data for the experiments in comp_trad_DL_meth must be requested from the organizers of the LeQua2024 quantification competition. Specifically, we requested the individual labels of dataset T2.