"""Model definitions used in the Sîmpetru et al. (2023 and 2024)"""
from functools import reduce
from typing import Any, Dict, Optional, Tuple, Union
import pytorch_lightning as pl
import torch
import torch.optim as optim
from torch import nn
[docs]
class RaulNetV9(pl.LightningModule):
"""Model definition used in Sîmpetru et al. [1]_ [2]_
Attributes
----------
learning_rate : float
The learning rate.
nr_of_input_channels : int
The number of input channels.
nr_of_outputs : int
The number of outputs.
cnn_encoder_channels : Tuple[int, int, int]
Tuple containing 3 integers defining the cnn encoder channels.
mlp_encoder_channels : Tuple[int, int]
Tuple containing 2 integers defining the mlp encoder channels.
event_search_kernel_length : int
Integer that sets the length of the kernels searching for action potentials.
event_search_kernel_stride : int
Integer that sets the stride of the kernels searching for action potentials.
Notes
-----
.. [1] Sîmpetru, R.C., März, M., Del Vecchio, A., 2023. Proportional and simultaneous real-time control of the full human hand from high-density electromyography. IEEE Trans. Neural Syst. Rehabil. Eng. 31, 3118–3131. https://doi.org/10/gsgk4s
.. [2] Sîmpetru, R.C., Cnejevici, V., Farina, D., Del Vecchio, A., 2024. Influence of spatio-temporal filtering on hand kinematics estimation from high-density EMG signals. J. Neural Eng. 21. https://doi.org/10/gtm4bt
"""
def __init__(
self,
learning_rate: float,
nr_of_input_channels: int,
input_length__samples: int,
nr_of_outputs: int,
cnn_encoder_channels: Tuple[int, int, int],
mlp_encoder_channels: Tuple[int, int],
event_search_kernel_length: int,
event_search_kernel_stride: int,
):
super(RaulNetV9, self).__init__()
self.save_hyperparameters()
self.learning_rate = learning_rate
self.nr_of_input_channels = nr_of_input_channels
self.nr_of_outputs = nr_of_outputs
self.input_length__samples = input_length__samples
self.cnn_encoder_channels = cnn_encoder_channels
self.mlp_encoder_channels = mlp_encoder_channels
self.event_search_kernel_length = event_search_kernel_length
self.event_search_kernel_stride = event_search_kernel_stride
self.criterion = nn.L1Loss()
self.cnn_encoder = nn.Sequential(
nn.Conv3d(
self.nr_of_input_channels,
self.cnn_encoder_channels[0],
kernel_size=(1, 1, self.event_search_kernel_length),
stride=(1, 1, self.event_search_kernel_stride),
groups=self.nr_of_input_channels,
),
nn.GELU(),
nn.InstanceNorm3d(self.cnn_encoder_channels[0], track_running_stats=False),
nn.Dropout3d(p=0.25),
nn.Conv3d(
self.cnn_encoder_channels[0],
self.cnn_encoder_channels[1],
kernel_size=(5, 32, 18),
dilation=(1, 2, 1),
padding=(2, 16, 0),
padding_mode="circular",
),
nn.GELU(),
nn.InstanceNorm3d(self.cnn_encoder_channels[1], track_running_stats=False),
nn.Conv3d(
self.cnn_encoder_channels[1],
self.cnn_encoder_channels[2],
kernel_size=(5, 9, 1),
),
nn.GELU(),
nn.InstanceNorm3d(self.cnn_encoder_channels[2], track_running_stats=False),
nn.Flatten(),
)
self.mlp_encoder = nn.Sequential(
nn.Linear(
reduce(
lambda x, y: x * int(y),
self.cnn_encoder(
torch.rand(
(
1,
self.nr_of_input_channels,
5,
64,
self.input_length__samples,
)
)
).shape[1:],
1,
),
self.mlp_encoder_channels[0],
),
nn.GELU(),
nn.Linear(self.mlp_encoder_channels[0], self.mlp_encoder_channels[1]),
nn.GELU(),
nn.Linear(self.mlp_encoder_channels[1], self.nr_of_outputs),
)
[docs]
def forward(self, inputs) -> torch.Tensor:
x = self._reshape_and_normalize(inputs)
x = self.cnn_encoder(x)
x = self.mlp_encoder(x)
return x
@staticmethod
def _reshape_and_normalize(inputs):
x = torch.stack(inputs.split(64, dim=2), dim=2)
return (x - x.mean(dim=(3, 4), keepdim=True)) / (
x.std(dim=(3, 4), keepdim=True, unbiased=True) + 1e-15
)
def configure_optimizers(self):
optimizer = optim.AdamW(
self.parameters(), lr=self.learning_rate, amsgrad=True, weight_decay=0.32
)
lr_scheduler = {
"scheduler": optim.lr_scheduler.OneCycleLR(
optimizer,
max_lr=self.learning_rate * (10**1.5),
total_steps=self.trainer.estimated_stepping_batches,
anneal_strategy="cos",
three_phase=False,
div_factor=10**1.5,
final_div_factor=1e3,
),
"name": "OneCycleLR",
"interval": "step",
"frequency": 1,
}
return [optimizer], [lr_scheduler]
def training_step(
self, train_batch, batch_idx: int
) -> Optional[Union[torch.Tensor, Dict[str, Any]]]:
inputs, ground_truths = train_batch
ground_truths = ground_truths[:, 0]
prediction = self(inputs)
scores_dict = {"loss": self.criterion(prediction, ground_truths)}
if scores_dict["loss"].isnan().item():
return None
self.log_dict(
scores_dict, prog_bar=True, logger=False, on_epoch=True, sync_dist=True
)
self.log_dict(
{f"train/{k}": v for k, v in scores_dict.items()},
prog_bar=False,
logger=True,
on_epoch=True,
on_step=False,
sync_dist=True,
)
return scores_dict
def validation_step(
self, batch, batch_idx
) -> Optional[Union[torch.Tensor, Dict[str, Any]]]:
inputs, ground_truths = batch
ground_truths = ground_truths[:, 0]
prediction = self(inputs)
scores_dict = {"val_loss": self.criterion(prediction, ground_truths)}
self.log_dict(
scores_dict, prog_bar=True, logger=False, on_epoch=True, sync_dist=True
)
return scores_dict
def test_step(
self, batch, batch_idx
) -> Optional[Union[torch.Tensor, Dict[str, Any]]]:
inputs, ground_truths = batch
ground_truths = ground_truths[:, 0]
prediction = self(inputs)
scores_dict = {"loss": self.criterion(prediction, ground_truths)}
self.log_dict(
scores_dict, prog_bar=True, logger=False, on_epoch=True, sync_dist=True
)
self.log_dict(
{f"test/{k}": v for k, v in scores_dict.items()},
prog_bar=False,
logger=True,
on_epoch=False,
on_step=True,
sync_dist=True,
)
return scores_dict