| { | |
| "model_class": "physioex.models.lseqsleepnet:LSeqSleepNet", | |
| "model_kwargs": { | |
| "n_classes": 5, | |
| "in_chan": 1, | |
| "F": 129, | |
| "D": 32, | |
| "nfft": 256, | |
| "sf": 100, | |
| "lowfreq": 0, | |
| "highfreq": 50, | |
| "epoch_hidden": 64, | |
| "epoch_attention": 64, | |
| "B": 10, | |
| "K": 20, | |
| "seq_hidden_ss": 64, | |
| "seq_hidden_ms": 64, | |
| "d_clf": 512, | |
| "dropout": 0.1 | |
| }, | |
| "training": { | |
| "dataset": "shhs", | |
| "dataset_kwargs": { | |
| "visit": 1 | |
| }, | |
| "channels": [ | |
| "EEG" | |
| ], | |
| "pipeline_preset": "seqsleepnet", | |
| "sequence_length": 200, | |
| "max_epochs": 50, | |
| "lr": 0.0001, | |
| "weight_decay": 0.0001, | |
| "batch_size": 8, | |
| "fold": 0 | |
| }, | |
| "reference": "Phan et al. 2023 - L-SeqSleepNet: Whole-cycle Long Sequence Modelling for Automatic Sleep Staging" | |
| } |