Networks

Нейронные сети и оптические системы


from svetlanna.networks import ...

Классы

LinearAutoencoder

Наследует: nn.Module

A simple autoencoder network consisting of consistent encoder and decoder for a simultaneous training. .encode() - forward propagation through encoder elements .decode() - forward propagation through decoder elements

Comment: Works only for a single wavelength, input wavefront [‘h’, ‘w’]

Свойства

device

Методы

__init__ constructor


__init__(self, sim_params: SimulationParameters, encoder_elements_list: list[Element] | Iterable[Element], decoder_elements_list: list[Element] | Iterable[Element], to_return: Literal['wf', 'amps'] = 'wf', device: str | torch.device = torch.get_default_device())

📥 Параметры

Параметр	Тип	Описание
`sim_params`	`SimulationParameters`	Simulation parameters for the task.
`encoder_elements_list`	`list[Element]	Iterable[Element]`
`decoder_elements_list`	`list[Element]	Iterable[Element]`
`to_return`	`Literal['wf', 'amps']`	Specifies what to return in .encode() and .decode(). (1) ‘wf’ – just a wavefront as it is (2) ‘amps’ – a wavefront without phases
`device`	`torch.device`	Device.

encode


encode(self, wavefront_in)

Propagation through the encoder part – encode an image wavefront (input).

📤 Возвращает

wavefront_encoded : Wavefront

An encoded input wavefront.

decode


decode(self, wavefront_encoded)

Propagation through the decoder part – decode an encoded image.

📤 Возвращает

wavefront_decoded : Wavefront

A decoded wavefront.

forward


forward(self, wavefront_in)

📥 Параметры

Параметр	Тип	Описание
`None`	`wavefront_in: Wavefront('bs', 'H', 'W')`

📤 Возвращает

wavefront_encoded, wavefront_decoded : torch.Wavefront

Encoded and decoded wavefronts.

to_specs


to_specs(self) -> Iterable[ParameterSpecs | SubelementSpecs]

to


to(self, device: str | torch.device | int) -> 'LinearAutoencoder'

ConvLayer4F

Наследует: nn.Module

Diffractive convolutional layer based on a 4f system.

Методы

__init__ constructor


__init__(self, sim_params: SimulationParameters, focal_length: float, conv_diffractive_mask: torch.Tensor, learnable_mask: bool = False, max_phase: float = 2 * torch.pi, fs_method: Literal['fresnel', 'AS'] = 'AS')

📥 Параметры

Параметр	Тип	Описание
`None`	`sim_params: SimulationParameters`	Simulation parameters for the task.
`None`	`focal_length: float`	A focal length for ThinLense’s in a 4f system.
`None`	`conv_diffractive_mask: torch.Tensor`	An initial mask for a DiffractiveLayer placed between two lenses in the system.
`None`	`learnable_mask: bool`	If True – a mask for a DiffractiveLayer in the 4f system will be learnable. Else – not learnable (const).
`None`	`max_phase: float`	A maximal phase for a Diffractive layer in the system.
`None`	`fs_method: Literal['fresnel', 'AS']`	A method for FreeSpace’s in the system.

get_free_space


get_free_space(self)

Returns a FreeSpace of a focal length for a 4f system.

get_thin_lens


get_thin_lens(self)

Returns a ThinLens with a pre-defined focal length.

get_diffractive_layer


get_diffractive_layer(self)

Returns a DiffractiveLayer according to pre-defined settings from init. It can be trainable or not according to self.learnable_mask flag.

get_conv_layer_4f


get_conv_layer_4f(self)

forward


forward(self, input_wf: Wavefront)

Forward propagation through a convolutional diffractive system based on a 4f system.

📥 Параметры

Параметр	Тип	Описание
`None`	`input_wf: Wavefront('batch_size', 'H', 'W')`	An input wavefront(s).

📤 Возвращает

: Wavefront

A wavefront after a propagation through a system.

ConvDiffNetwork4F

Наследует: nn.Module

A simple convolutional network with a 4f system as an optical convolutional layer. Comment: -> [4f system (convolution)] -> [some system of elements] ->

Свойства

device

Методы

__init__ constructor


__init__(self, sim_params: SimulationParameters, network_elements_list: list, focal_length: float, conv_phase_mask: torch.Tensor, learnable_mask: bool = False, max_phase: float = 2 * torch.pi, fs_method: Literal['fresnel', 'AS'] = 'AS', device: str | torch.device = torch.get_default_device())

📥 Параметры

Параметр	Тип	Описание
`sim_params`	`SimulationParameters`	Simulation parameters for the task.
`network_elements_list`	`list`	List of Elements for a Network after a convolutional layer (4f system).
`None`	`focal_length: float`	A focal length for ThinLense’s in a 4f system.
`None`	`conv_phase_mask: torch.Tensor`	An initial mask for a DiffractiveLayer placed between two lenses in the system.
`None`	`learnable_mask: bool`	If True – a mask for a DiffractiveLayer will be learnable. Else – not learnable (const).
`None`	`max_phase: float`	A maximal phase for a Diffractive layer in the 4f system.
`None`	`fs_method: Literal['fresnel', 'AS']`	A method for FreeSpace’s in the system.
`None`	`device: str	torch.device`

forward


forward(self, wavefront_in)

📥 Параметры

Параметр	Тип	Описание
`None`	`wavefront_in: Wavefront('bs', 'H', 'W')`	Input wavefront or a batch of Wavefronts.

📤 Возвращает

: torch.Tensor | Wavefront

Output after a Wavefront propagation through a Convolutional layer and the Other Part of all Network.

to_specs


to_specs(self) -> Iterable[ParameterSpecs | SubelementSpecs]

to


to(self, device: str | torch.device | int) -> 'ConvDiffNetwork4F'

DiffractiveRNN

Наследует: nn.Module

A simple recurrent diffractive network of an architecture proposed in the article: https://www.nature.com/articles/s41566-021-00796-w

Свойства

device

Методы

__init__ constructor


__init__(self, sim_params: SimulationParameters, sequence_len: int, fusing_coeff: float, read_in_layer: nn.Sequential, memory_layer: nn.Sequential, hidden_forward_layer: nn.Sequential, read_out_layer: nn.Sequential, detector_layer: nn.Sequential, device: str | torch.device = torch.get_default_device())

sim_params: SimulationParameters Simulation parameters for the task. sequence_len: int A size (number of frames) of sequences (of Wavefronts) for prediction. fusing_coeff: float A coefficient in a function for a hidden state (lambda in methods of the article). read_in_layer, memory_layer: nn.Sequential Systems of elements for a D-RNN parts (see the article). hidden_forward_layer: nn.Sequential System for a hidden state after each frame input. Comment: mix_i = (1 - fusing_coeff) * read_in_layer(input_i) + fusing_coeff * hidden_i hidden_i = hidden_forward_layer(mix_i) read_out_layer, detector_layer: nn.Sequential System of elements for a D-RNN output. detector_layer ends with a Detector! device: torch.device Specified device.

forward


forward(self, subsequence_wf: Wavefront)

📥 Параметры

Параметр	Тип	Описание
`None`	`subsequence_wf: Wavefront('batch_size', 'sequence_len', 'H', 'W')`	Wavefronts for a sequence. Comment: works for a single wavelength in SimulationParameters!

to_specs


to_specs(self) -> Iterable[ParameterSpecs | SubelementSpecs]

to


to(self, device: str | torch.device | int) -> 'DiffractiveRNN'