-
Notifications
You must be signed in to change notification settings - Fork 1
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request #63 from aertslab/main
Enhancer design motif implementation fix
- Loading branch information
Showing
7 changed files
with
435 additions
and
74 deletions.
There are no files selected for viewing
Large diffs are not rendered by default.
Oops, something went wrong.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,122 @@ | ||
| import keras | ||
| import keras.ops as ops | ||
|
|
||
|
|
||
| @keras.saving.register_keras_serializable(package="Losses") | ||
| class PoissonMultinomialLoss(keras.losses.Loss): | ||
| """ | ||
| Poisson decomposition with multinomial specificity term for aggregated counts. | ||
| Combines Poisson loss for total counts with a multinomial term for class proportions. | ||
| Parameters | ||
| ---------- | ||
| total_weight : float | ||
| Weight of the Poisson term in the total loss. | ||
| eps : float | ||
| Small value to avoid log(0). | ||
| log_input : bool | ||
| If True, applies exponential transformation to predictions to produce counts. | ||
| multinomial_axis : str | ||
| Either "length" or "task", representing the axis along which multinomial proportions are calculated. | ||
| reduction : str | ||
| Type of reduction to apply to the loss: "mean" or "none". | ||
| name : str | ||
| Name of the loss function. | ||
| """ | ||
|
|
||
| def __init__( | ||
| self, | ||
| total_weight: float = 1.0, | ||
| eps: float = 1e-7, | ||
| log_input: bool = True, | ||
| multinomial_axis: str = "task", | ||
| reduction: str = "sum_over_batch_size", | ||
| name: str = "PoissonMultinomialLoss", | ||
| ): | ||
| """ | ||
| Initialize the PoissonMultinomialLoss. | ||
| Parameters | ||
| ---------- | ||
| total_weight : float, optional | ||
| Weight of the Poisson term in the total loss (default is 1.0). | ||
| eps : float, optional | ||
| Small value to avoid log(0) (default is 1e-7). | ||
| log_input : bool, optional | ||
| If True, applies exponential transformation to predictions to produce counts (default is True). | ||
| multinomial_axis : str, optional | ||
| Either "length" or "task", representing the axis along which multinomial proportions are calculated | ||
| (default is "task"). | ||
| reduction : str, optional | ||
| Type of reduction to apply to the loss: "mean" or "none" (default is "sum_over_batch_size"). | ||
| name : str, optional | ||
| Name of the loss function (default is "PoissonMultinomialLoss"). | ||
| """ | ||
| super().__init__(name=name, reduction=reduction) | ||
| self.total_weight = total_weight | ||
| self.eps = eps | ||
| self.log_input = log_input | ||
| self.axis = 1 if multinomial_axis == "task" else 0 | ||
|
|
||
| def call(self, y_true, y_pred): | ||
| """ | ||
| Compute the PoissonMultinomialLoss. | ||
| Parameters | ||
| ---------- | ||
| y_true : Tensor | ||
| True target values (aggregated counts). | ||
| y_pred : Tensor | ||
| Predicted values. | ||
| Returns | ||
| ------- | ||
| Tensor | ||
| Combined loss value. | ||
| """ | ||
| # Ensure predictions and targets are float32 | ||
| y_true = ops.cast(y_true, dtype="float32") | ||
| y_pred = ops.cast(y_pred, dtype="float32") | ||
|
|
||
| # Apply exp if log_input is True | ||
| if self.log_input: | ||
| y_pred = ops.log(y_pred+1) | ||
| y_true = ops.log(y_true+1) | ||
|
|
||
| # Total counts along the specified axis | ||
| total_true = ops.sum(y_true, axis=self.axis, keepdims=True) | ||
| total_pred = ops.sum(y_pred, axis=self.axis, keepdims=True) | ||
|
|
||
| # Poisson term | ||
| poisson_term = ( | ||
| total_pred - total_true * ops.log(total_pred + self.eps) | ||
| ) | ||
|
|
||
| # Multinomial probabilities | ||
| p_pred = y_pred / (total_pred + self.eps) | ||
| log_p_pred = ops.log(p_pred + self.eps) | ||
|
|
||
| # Multinomial term | ||
| multinomial_dot = -y_true * log_p_pred | ||
| multinomial_term = ops.sum(multinomial_dot, axis=self.axis, keepdims=True) | ||
|
|
||
| # Combine Poisson and Multinomial terms | ||
| loss = multinomial_term + self.total_weight * poisson_term | ||
|
|
||
| # Apply reduction | ||
| if self.reduction == "mean": | ||
| return ops.mean(loss) | ||
| else: | ||
| return loss | ||
|
|
||
| def get_config(self): | ||
| """Return the configuration of the loss function.""" | ||
| config = super().get_config() | ||
| config.update({ | ||
| "total_weight": self.total_weight, | ||
| "eps": self.eps, | ||
| "log_input": self.log_input, | ||
| "axis": self.axis, | ||
| }) | ||
| return config |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.