matchclot.utils.dataloaders module

Summary

Classes:

AugDataset	Modality matching dataset class for training with affine transformation data augmentations.
ModalityMatchingDataset
SingleAugDataset	Similar to AugDataset class, but only applies a single affine transformation to each modality.

Functions:

get_dataloaders

Reference

class ModalityMatchingDataset(df_modality1, df_modality2)

Bases: Dataset

get_dataloaders(mod1_train, mod2_train, sol_train, mod1_test, mod2_test, sol_test, NUM_WORKERS, BATCH_SIZE)

class AugDataset(df_modality1, df_modality2, transf_matrices_mod1, transf_vectors_mod1, transf_matrices_mod2, transf_vectors_mod2)

Bases: Dataset

Modality matching dataset class for training with affine transformation data augmentations. Given 4 affine transformations for each modality, it applies a random transformation which is an interpolation of the 4 transformations. Can be used to augment a distribution of cell profiles from a single batch by applying affine transformations that approximate the optimal transport map between the non-augmented batch and other batches.

__init__(df_modality1, df_modality2, transf_matrices_mod1, transf_vectors_mod1, transf_matrices_mod2, transf_vectors_mod2)

Parameters:

• df_modality1 – pandas dataframe of mod1 data

• df_modality2 – pandas dataframe of mod2 data

• transf_matrices_mod1 – list of 4 linear transformation matrices for augmenting mod1

• transf_vectors_mod1 – list of 4 translation vectors for augmenting mod1

• transf_matrices_mod2 – list of 4 linear transformation matrices for augmenting mod2

• transf_vectors_mod2 – list of 4 translation vectors for augmenting mod2

class SingleAugDataset(df_modality1, df_modality2, transf_matrix_mod1, transf_vector_mod1, transf_matrix_mod2, transf_vector_mod2)

Bases: Dataset

Similar to AugDataset class, but only applies a single affine transformation to each modality.

__init__(df_modality1, df_modality2, transf_matrix_mod1, transf_vector_mod1, transf_matrix_mod2, transf_vector_mod2)

Parameters:

• df_modality1 – pandas dataframe of mod1 data

• df_modality2 – pandas dataframe of mod2 data

• transf_matrices_mod1 – linear transformation matrix for augmenting mod1

• transf_vectors_mod1 – translation vector for augmenting mod1

• transf_matrices_mod2 – linear transformation matrix for augmenting mod2

• transf_vectors_mod2 – translation vector for augmenting mod2