Data

class torchmimic.data.DecompensationDataset(root, train=True, n_samples=None)

Decompensation dataset that can be directly used by PyTorch dataloaders. This class preprocessing the data the same way as “Multitask learning and benchmarking with clinical time series data”: https://github.com/YerevaNN/mimic3-benchmarks

Parameters

• root (str) – directory where data is located

• train (bool) – if true, the training split of the data will be used. Otherwise, the validation dataset will be used

• n_samples – number of samples to use. If None, all the data is used

class torchmimic.data.IHMDataset(root, train=True, n_samples=None)

In-Hospital-Mortality dataset that can be directly used by PyTorch dataloaders. This class preprocessing the data the same way as “Multitask learning and benchmarking with clinical time series data”: https://github.com/YerevaNN/mimic3-benchmarks

Parameters

• root (str) – directory where data is located

• train (bool) – if true, the training split of the data will be used. Otherwise, the validation dataset will be used

• n_samples – number of samples to use. If None, all the data is used

class torchmimic.data.LOSDataset(root, train=True, partition=10, n_samples=None)

Length-of-Stay dataset that can be directly used by PyTorch dataloaders. This class preprocessing the data the same way as “Multitask learning and benchmarking with clinical time series data”: https://github.com/YerevaNN/mimic3-benchmarks

Parameters

• root (str) – directory where data is located

• train (bool) – if true, the training split of the data will be used. Otherwise, the validation dataset will be used

• partition – number of patitions to use for binning

• n_samples – number of samples to use. If None, all the data is used

class torchmimic.data.PhenotypingDataset(root, train=True, n_samples=None)

Phenotyping dataset that can be directly used by PyTorch dataloaders. This class preprocessing the data the same way as “Multitask learning and benchmarking with clinical time series data”: https://github.com/YerevaNN/mimic3-benchmarks

Parameters

• root (str) – directory where data is located

• train (bool) – if true, the training split of the data will be used. Otherwise, the validation dataset will be used

• n_samples – number of samples to use. If None, all the data is used