LALossTrainer

class torcheeg.trainers.LALossTrainer(model: Module, num_classes: int, class_frequency: List[int] | DataLoader, tau: float = 1.0, eps: float = 1e-12, lr: float = 0.001, weight_decay: float = 0.0, devices: int = 1, accelerator: str = 'cpu', metrics: List[str] = ['accuracy'])

A trainer class for EEG classification with Logit-adjusted (LA) loss for imbalanced datasets.

• Paper: Menon A K, Jayasumana S, Rawat A S, et al. Long-tail learning via logit adjustment[J]. arXiv preprint arXiv:2007.07314, 2020.

• URL: https://arxiv.org/abs/2007.07314

• Related Project: https://github.com/Chumsy0725/logit-adj-pytorch

from torcheeg.models import CCNN
from torcheeg.trainers import LALossTrainer

model = CCNN(in_channels=5, num_classes=2)
trainer = LALossTrainer(model, num_classes=2, class_frequency=[10, 20], tau=1.0)

Parameters:

• model (nn.Module) – The classification model, and the dimension of its output should be equal to the number of categories in the dataset. The output layer does not need to have a softmax activation function.

• num_classes (int) – The number of classes in the dataset.

• class_frequency (List[int] or Dataloader) – The frequency of each class in the dataset. It can be a list of integers or a dataloader to calculate the frequency of each class in the dataset, traversing the data batch (torch.utils.data.dataloader.DataLoader, torch_geometric.loader.DataLoader, etc). (default: None)

• tau (float) – The temperature parameter. (default: 1.0)

• eps (float) – The epsilon parameter. (default: 1e-12)

• lr (float) – The learning rate. (default: 0.001)

• weight_decay (float) – The weight decay. (default: 0.0)

• devices (int) – The number of devices to use. (default: 1)

• accelerator (str) – The accelerator to use. Availabel options are: ‘cpu’, ‘gpu’. (default: "cpu")

• metrics (list of str) – The metrics to use. Availabel options are: ‘precision’, ‘recall’, ‘f1score’, ‘accuracy’, ‘matthews’, ‘auroc’, and ‘kappa’. (default: ["accuracy"])

fit(train_loader: DataLoader, val_loader: DataLoader, max_epochs: int = 300, *args, **kwargs) → Any

Parameters:

• train_loader (DataLoader) – Iterable DataLoader for traversing the training data batch (torch.utils.data.dataloader.DataLoader, torch_geometric.loader.DataLoader, etc).

• val_loader (DataLoader) – Iterable DataLoader for traversing the validation data batch (torch.utils.data.dataloader.DataLoader, torch_geometric.loader.DataLoader, etc).

• max_epochs (int) – Maximum number of epochs to train the model. (default: 300)

test(test_loader: DataLoader, *args, **kwargs) → List[Dict[str, float]]

Parameters:

test_loader (DataLoader) – Iterable DataLoader for traversing the test data batch (torch.utils.data.dataloader.DataLoader, torch_geometric.loader.DataLoader, etc).