Class EDMDDLSolver
The class EDMDDLSolver implements the EDMD-DL algorithm.
It's a subclass of EDMDSolver.
API Documentation¶
Bases: EDMDSolver
__init__(dictionary, reg, reg_final=0.01)
¶
Initialize the EDMDDLSolver instance.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dictionary
|
TrainableDictionary
|
The dictionary used in the algorithm. |
required |
reg
|
float
|
The regularization parameter used in the algorithm. |
required |
reg_final
|
float
|
The final regularization parameter used in the algorithm. |
0.01
|
solve(dataset_train, dataset_val, n_epochs, batch_size, tol=1e-08, lr=0.0001)
¶
Solves the Koopman operator learning problem using the provided training and validation datasets.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dataset_train
|
KoopmanDataset
|
The training dataset containing input data and labels. |
required |
dataset_val
|
KoopmanDataset
|
The validation dataset containing input data and labels. |
required |
n_epochs
|
int
|
The number of epochs to train the model. |
required |
batch_size
|
int
|
The size of each batch for training and validation. |
required |
tol
|
float
|
The tolerance for early stopping based on training loss. Defaults to 1e-8. |
1e-08
|
lr
|
float
|
The learning rate for the optimizer. Defaults to 1e-4. |
0.0001
|
Returns:
| Type | Description |
|---|---|
Koopman
|
The learned Koopman operator. |