snrv.Snrv

class snrv.Snrv(input_size, output_size, hidden_depth=2, hidden_size=100, activation=ReLU(), batch_norm=False, dropout_rate=0.0, lr=0.1, weight_decay=0.0, val_frac=0.2, n_epochs=100, batch_size=100, VAMPdegree=2, is_reversible=True)

ANN encoder for state-free reversible VAMPnet

Parameters

• input_size (int) – number of neurons in input layer

• output_size (int) – number of neurons in output layer corresponding to number of basis functions to compute for VAC / VAMP

• hidden_depth (int, default = 2) – number of hidden layers

• hidden_size (int, default = 100) – number of neurons in each hidden layer

• activation (str, default = nn.ReLU()) – activation function to be used in each neuron

• batch_norm (bool, default = False) – use batch normalization after each layer

• dropout_rate (float, default = 0.) – dropout rate to use in each layer

• lr (float, default = 0.1) – learning rate for Adam optimizer

• weight_decay (float, default = 0.) – weight decay for Adam optimizer

• val_frac (float, default = 0.2) – fraction of data to place in validation partition; balance is used for training

• n_epochs (int, default = 100) – number of training epochs for ANN

• batch_size (int, default = 100) – no. of instances per mini batch

• VAMPdegree (int, default = 2) – exponent to use in VAMP-r score loss function

• is_reversible (bool, default = True) – indicator flag as to whether to enforce detailed balance by symmetrizing trajectory by augmenting with time reversed twin and solve VAC OR not assume microscopic reversibility and solve VAMP

Variables

• self.device (str) – specification as to whether to use ‘cpu’ or ‘cuda’ for model training

• self.is_fitted (bool) – indicator flag as to whether or nor model has been fitted

• self._train_loader (DataLoader object) – training data loader

• self._val_loader (DataLoader object) – validation data loader

• self.lag (int) – lag in steps to apply to data trajectory

• self.optimizer (torch.optim.Adam object) – optimizer for backpropagation

• self._train_step (function) – training step function produced by self._make_train_step

• self.evals (torch.tensor, n_comp, n_comp = no. of basis functions in ANN == output_size) – eigenvalues of VAC generalized eigenvalue problem finding linear combination of learned basis functions to produce approximations of transfer operator eigenvectors in non-ascending order OR singular values of VAMP singular value problem finding linear combination of learned basis functions to produce approximations of transfer operator left and right singular vectors in non-ascending order

• self.expansion_coefficients (n_comp, n_comp = no. of basis functions in ANN == output_size) – expansion coefficients for linear combination of learned basis functions into transfer operator eigenvectors (reversible) or left singular vectors (non-reversible)

• self.expansion_coefficients_right (n_comp, n_comp = no. of basis functions in ANN == output_size) – expansion coefficients for linear combination of learned basis functions into transfer operator right singular vectors (non-reversible)

• self.training_losses (list, n_epoch) – loss over training data in each epoch

• self.validation_losses (list, n_epoch) – loss over validation data in each epoch

__init__(input_size, output_size, hidden_depth=2, hidden_size=100, activation=ReLU(), batch_norm=False, dropout_rate=0.0, lr=0.1, weight_decay=0.0, val_frac=0.2, n_epochs=100, batch_size=100, VAMPdegree=2, is_reversible=True)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

Methods

__init__(input_size, output_size[, ...])	Initializes internal Module state, shared by both nn.Module and ScriptModule.
add_module(name, module)	Adds a child module to the current module.
apply(fn)	Applies fn recursively to every submodule (as returned by .children()) as well as self.
bfloat16()	Casts all floating point parameters and buffers to bfloat16 datatype.
buffers([recurse])	Returns an iterator over module buffers.
children()	Returns an iterator over immediate children modules.
cpu()	Moves all model parameters and buffers to the CPU.
cuda([device])	Moves all model parameters and buffers to the GPU.
double()	Casts all floating point parameters and buffers to double datatype.
eval()	Sets the module in evaluation mode.
extra_repr()	Set the extra representation of the module
fit(data, lag[, ln_pathweight])	fit SNRV model to data
fit_transform(data, lag[, ln_pathweight])	fit SNRV over data then project data into learned eigenvector (VAC) / singular vectors (VAMP) of transfer operator
float()	Casts all floating point parameters and buffers to float datatype.
forward(x_t0, x_tt)	Forward pass through network
get_transform_Jacobian(data)	compute Jacobian of self.transform computational graph output (psi) wrt input (data)
half()	Casts all floating point parameters and buffers to half datatype.
load_state_dict(state_dict[, strict])	Copies parameters and buffers from state_dict into this module and its descendants.
load_weights(network_weights)	loading SNRV network weights from file
modules()	Returns an iterator over all modules in the network.
named_buffers([prefix, recurse])	Returns an iterator over module buffers, yielding both the name of the buffer as well as the buffer itself.
named_children()	Returns an iterator over immediate children modules, yielding both the name of the module as well as the module itself.
named_modules([memo, prefix])	Returns an iterator over all modules in the network, yielding both the name of the module as well as the module itself.
named_parameters([prefix, recurse])	Returns an iterator over module parameters, yielding both the name of the parameter as well as the parameter itself.
parameters([recurse])	Returns an iterator over module parameters.
register_backward_hook(hook)	Registers a backward hook on the module.
register_buffer(name, tensor[, persistent])	Adds a buffer to the module.
register_forward_hook(hook)	Registers a forward hook on the module.
register_forward_pre_hook(hook)	Registers a forward pre-hook on the module.
register_full_backward_hook(hook)	Registers a backward hook on the module.
register_parameter(name, param)	Adds a parameter to the module.
requires_grad_([requires_grad])	Change if autograd should record operations on parameters in this module.
save_model(modelFilePath)	saving model parameters required for building and running (i.e., self.transform)
share_memory()
state_dict([destination, prefix, keep_vars])	Returns a dictionary containing a whole state of the module.
to(*args, **kwargs)	Moves and/or casts the parameters and buffers.
train([mode])	Sets the module in training mode.
transform(data)	project data into learned eigenvector basis
type(dst_type)	Casts all parameters and buffers to dst_type.
xpu([device])	Moves all model parameters and buffers to the XPU.
zero_grad([set_to_none])	Sets gradients of all model parameters to zero.

Attributes

T_destination	alias of TypeVar('T_destination', bound=Mapping[str, torch.Tensor])
dump_patches	This allows better BC support for load_state_dict().