examples.systems.parameters.nanorings

@dataclass(frozen=True)
class NanoringsHyperparameters:

Hyperparameters for the Nanorings digital twin experiment.

NanoringsHyperparameters( DRIFT_LEARNING_RATE: float = 0.003, DIFFUSION_LEARNING_RATE: float = 0.0005, GENERATOR_LEARNING_RATE: float = 0.0001, CRITIC_LEARNING_RATE: float = 0.0005, NUMBER_OF_EPOCHS: int = 1024, BATCH_SIZE: int = 128, GAN_BATCH_SIZE: int = 128, VALIDATION_SPLIT: float = 0.2, EARLY_STOPPING_PATIENCE: int = 150, NUMBER_OF_GAN_EPOCHS: int = 400, CRITIC_WINDOW_SIZE: int = 64, NUMBER_OF_SDE_ROLLOUT_SAMPLES: int = 256, MOMENT_MATCHING_WEIGHT: float = 100.0, MOMENT_MATCHING_ENABLED: bool = True, CRITIC_UPDATES: int = 1, SDE_L1_WEIGHT: float = 2.0, DRIFT_L1_WEIGHT: float = 1.0, GRADIENT_PENALTY_WEIGHT: float = 10.0, TRAIN_ODE_WITH_SDE: bool = True, PERCENTAGE_OF_FILES_TO_LOAD: float = 1.0, PERCENTAGE_OF_MAIN_SEQUENCE_TO_USE: float = 0.05, TARGET_CHANNEL: int = 0, SIGNALS: int = 1, SAMPLE_RATE: int = 3200, H_FIELD_AMPLITUDE_UPDATE_RATE: int = 100, TRANSIENT_LENGTH: int = 300, CONTEXT_POINTS: int = 0, DRIFT_NET_ARCH: neural_dynamics.core.hyperparameters.NetworkArchitecture = NetworkArchitecture(input_size=4, hidden_sizes=[128, 128, 128, 128], output_size=1), DIFFUSION_NET_ARCH: neural_dynamics.core.hyperparameters.NetworkArchitecture = NetworkArchitecture(input_size=4, hidden_sizes=[64, 64], output_size=1), CRITIC_NET_ARCH: neural_dynamics.core.hyperparameters.NetworkArchitecture = NetworkArchitecture(input_size=64, hidden_sizes=[128, 128, 64], output_size=1))
DRIFT_LEARNING_RATE: float = 0.003
DIFFUSION_LEARNING_RATE: float = 0.0005
GENERATOR_LEARNING_RATE: float = 0.0001
CRITIC_LEARNING_RATE: float = 0.0005
NUMBER_OF_EPOCHS: int = 1024
BATCH_SIZE: int = 128
GAN_BATCH_SIZE: int = 128
VALIDATION_SPLIT: float = 0.2
EARLY_STOPPING_PATIENCE: int = 150
NUMBER_OF_GAN_EPOCHS: int = 400
CRITIC_WINDOW_SIZE: int = 64
NUMBER_OF_SDE_ROLLOUT_SAMPLES: int = 256
MOMENT_MATCHING_WEIGHT: float = 100.0
MOMENT_MATCHING_ENABLED: bool = True
CRITIC_UPDATES: int = 1
SDE_L1_WEIGHT: float = 2.0
DRIFT_L1_WEIGHT: float = 1.0
GRADIENT_PENALTY_WEIGHT: float = 10.0
TRAIN_ODE_WITH_SDE: bool = True
PERCENTAGE_OF_FILES_TO_LOAD: float = 1.0
PERCENTAGE_OF_MAIN_SEQUENCE_TO_USE: float = 0.05
TARGET_CHANNEL: int = 0
SIGNALS: int = 1
SAMPLE_RATE: int = 3200
H_FIELD_AMPLITUDE_UPDATE_RATE: int = 100
TRANSIENT_LENGTH: int = 300
CONTEXT_POINTS: int = 0
DRIFT_NET_ARCH: neural_dynamics.core.hyperparameters.NetworkArchitecture = NetworkArchitecture(input_size=4, hidden_sizes=[128, 128, 128, 128], output_size=1)
DIFFUSION_NET_ARCH: neural_dynamics.core.hyperparameters.NetworkArchitecture = NetworkArchitecture(input_size=4, hidden_sizes=[64, 64], output_size=1)
CRITIC_NET_ARCH: neural_dynamics.core.hyperparameters.NetworkArchitecture = NetworkArchitecture(input_size=64, hidden_sizes=[128, 128, 64], output_size=1)
def to_hyperparameters(self) -> neural_dynamics.core.hyperparameters.Hyperparameters:

Convert to standard Hyperparameters object for Neural SDE training.