Generative model for inverse design of materials, devices, and structures

Patent metadata

Field	Value
Publication number	US-12260339-B2
Application number	US-202117306003-A
Country	US
Kind code	B2
Filing date	May 3, 2021
Priority date	May 3, 2021
Publication date	Mar 25, 2025
Grant date	Mar 25, 2025

Abstract

A photonic device for splitting optical beams includes an input port configured to receive an input beam having an input power, a power splitter including perturbation segments arranged in a first region and a second region of a guide material having a first refractive index, each segment having a second refractive index, wherein the first region is configured to split the input beam into a first beam and a second beam, wherein and the second region is configured to separately guide the first and second beams, wherein the first refractive index is greater than the second refractive index, and output ports including first and second output ports connected the power splitter to respectively receive and transmit the first and second beams.

First claim

The invention claimed is:

1. A system for training a device design network for generating layouts of photonic devices, comprising:
an interface configured to acquire input data of a photonic device, wherein the input data includes user-desired transmission information given as a condition, and a Gaussian distribution;
a memory storing the device design network including a first encoder, a second encoder, a first decoder, a second decoder, and a first adversarial block and a second adversarial block, wherein the first and second encoders have a same structure and share same weights, the first and second decoders have a same structure and share same weights, and the first and second adversarial blocks have a same structure and share same weights; and
a processor, in connection with the memory, configured to:

train the device design network using training data including an input pattern, wherein the input pattern is a hole vector pattern, and an input condition, wherein the input condition is a transmission efficiency;
update the weights of the first and second encoders and the first and second decoders based on a sum of a first loss function and a third loss function to reduce a difference between the input pattern and output pattern of the first and second decoders; and
update the weights in the first and second adversarial blocks by minimizing a second loss function, wherein the second loss function maximizes a loss between the condition and the first and second adversarial conditions.

2. The system of claim 1 , wherein the first and second encoders are constructed by at least one convolutional layer followed by at least one parallel fully connected layer to extract features of a layout of the photonic device.

3. The system of claim 2 , wherein each of the at least one convolutional layer includes more than two channels.

4. The system of claim 2 , wherein each of the at least one parallel fully connected layer includes two input/output dimensions.

5. The system of claim 1 , wherein the photonic device is an optical power splitter, wherein the extracted device features are mean (μ) and covariance (σ) for the Gaussian distribution.

6. The system of claim 1 , wherein the photonic device is a power splitter.

7. The system of claim 1 , wherein the photonic device is a wavelength splitter.

8. The system of claim 1 , wherein the photonic device is a mode converter.

9. The system of claim 1 , wherein the training data comprises of device structures optimized by an adjoint method.

10. The system of claim 3 , wherein the two convolutional layers include 8 channels and 16 channels, respectively.

11. The system of claim 4 , wherein each of the two parallel fully connected layer include 800 input/output dimensions and 60 input/output dimensions, respectively.

12. A computer-implemented training method for training a device design network, wherein the method comprising steps of:
acquire input data of a photonic device via an interface, wherein the input data includes user-desired transmission information given as a condition, and a Gaussian distribution;
train the device design network using training data including an input pattern, wherein the input pattern is a hole vector pattern, and an input condition, wherein the input condition is a transmission efficiency;

update the weights of the first and second encoders and the first and second decoders based on a sum of a first loss function and a third loss function to reduce a difference between the input pattern and output pattern of the first encoder and the second encoder and the first decoder and the second decoders;
update the weights in a first adversarial block and a second adversarial block by minimizing a second loss function, wherein the second loss function maximizes a loss between the condition and the first and second adversarial conditions; and
store the device design network including a first encoder, a second encoder, a first decoder, a second decoder, and a first adversarial block and a second adversarial block, wherein the first and second encoders have a same structure and share same weights, the first and second decoders have a same structure and share same weights, and the first and second adversarial blocks have a same structure and share same weights.

13. The method of claim 12 , wherein the first and second encoders are constructed by at least one convolutional layer followed by at least one parallel fully connected layer to extract features of a layout of the photonic device.

14. The method of claim 13 , wherein each of the at least one convolutional layer includes more than two channels.

15. The method of claim 13 , wherein each of the at least one parallel fully connected layer includes two input/output dimensions.

16. The method of claim 12 , wherein the photonic device is an optical power splitter, wherein the extracted device features are mean (μ) and covariance (σ) for the Gaussian distribution.

17. The method of claim 12 , wherein the first loss function is expressed by combination of BCE Loss of between the input and output of the first encoder and decoder set and the KL-Divergence between the encoded latent, the standard Gaussian Distribution, the encoded latent and the output of the first adversarial block, wherein the second loss function is expressed by combination of Mean Square Root Loss (MSE Loss) between the encoded latent and the output from the first adversarial block and the MSE loss between the condition between the encoded latent and the output from the second adversarial block, wherein the third loss function is expressed by combination of the MSE Loss between the standard gaussian samples and the second latent variables, and the MSE loss between the encoded latent and the output of the adversarial block.

18. A computer-implemented method for generating layouts of photonic devices using a device generating network, comprising steps:
acquiring input data of the photonic device via an interface, wherein the input data includes user-desired transmission information given as a condition, and a Gaussian distribution;
feeding the input data into the device generating network, wherein the device generating network is pretrained by a computer-implemented training method of claim 12 ; and
generating layout data of a layout of the photonic device using the pretrained device generating network and storing the layout data into a memory.

19. The method of claim 18 , wherein the photonic device is a mode converter.

20. The method of claim 18 , wherein the training data comprises of device structures optimized by an adjoint method.

Assignees

• Mitsubishi Electric Res Laboratories Inc

Inventors

• Kojima Keisuke
• Koike Akino Toshiaki
• Tang Yingheng
• Wang Ye

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/091

  Active learning · CPC title

• G06N3/094

  Adversarial learning · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title