Epistemic and Aleatoric Deep Plasticity based on Sound Feedback
US-2019279094-A1 · Sep 12, 2019 · US
US11636241B2 · US · B2
| Field | Value |
|---|---|
| Publication number | US-11636241-B2 |
| Application number | US-202117387082-A |
| Country | US |
| Kind code | B2 |
| Filing date | Jul 28, 2021 |
| Priority date | Apr 29, 2019 |
| Publication date | Apr 25, 2023 |
| Grant date | Apr 25, 2023 |
A practical reading order for non-experts. Skip the full description unless you need deep technical detail.
What the patent document calls the invention.
A short plain-language summary of the technical disclosure.
Who owns or filed the patent and who is credited as inventor.
Filing, priority, publication, and grant dates set the timeline.
The legal scope of protection — read this for what is actually claimed.
Technology tags used to group this patent with similar filings.
Prior art links and similar publications in this corpus.
Official abstract text for this publication.
A system, apparatus, and method for optimizing structural parameters of a physical device are described. The method includes receiving an initial description of the physical device describing the structural parameters within a simulated environment. The method further includes performing a simulation of the physical device in response to an excitation source to determine a performance metric of the physical device. The simulation environment includes one or more absorbing boundaries for attenuation of an output of the excitation source during the simulation. The method further includes recording attenuated field values of the simulated environment associated with the attenuation during the simulation. The method further includes determining a loss metric based on a difference between the performance metric and a target performance metric, backpropagating the loss metric using the attenuated field values, and generating a revised description of the physical device by updating the structural parameters to reduce the loss metric.
Opening claim text (preview).
What is claimed is: 1. A computer-implemented method, comprising: receiving an initial description of a physical device represented by structural parameters within a simulated environment, performing a simulation of the physical device in response to an excitation source over a plurality of time steps, wherein the simulated environment includes one or more absorbing boundaries for attenuation of an output of the excitation source during the simulation, wherein the one or more absorbing boundaries surround a plurality of voxels representative of the physical device; recording attenuated field values of the simulated environment associated with the attenuation during the simulation over the plurality of time steps of the simulation; and performing a reverse simulation of the physical device in which the attenuated field values are replayed over the plurality of time steps in reverse to recover a field response of the simulated environment at one or more of the plurality of time steps for one or more voxels included in the plurality of voxels representative of the physical device. 2. The computer-implemented method of claim 1 , wherein the physical device is an electromagnetic device, wherein the output of the excitation source corresponds to electromagnetic radiation, and wherein the attenuated field values correspond to at least one of electric field or magnetic field at the one or more absorbing boundaries within the simulated environment. 3. The computer-implemented method of claim 1 , wherein during the reverse simulation the attenuated field values correspond to one or more excitation sources of the reverse simulation that originate at the one or more absorbing boundaries of the simulated environment, and wherein the attenuated field values are replayed from a particular time step included in the plurality of time steps to an earlier time step included in the plurality of time steps. 4. The computer-implemented method of claim 1 , wherein the attenuated field values recorded over the plurality of time steps provide a reduced memory footprint relative to a recordation of the field response of the simulated environment during the simulation. 5. The computer-implemented method of claim 1 , wherein the simulated environment includes the plurality of voxels that collectively describe the structural parameters of the physical device, wherein each of the plurality of voxels is associated with a structural value to describe the structural parameters, a field value to describe a field response to the excitation source, and a source value to describe the excitation source, wherein the field response includes the attenuated field values. 6. The computer-implemented method of claim 5 , wherein the attenuated field values correspond to the field value of a portion of the plurality of voxels proximate to the absorbing boundaries, and wherein the output of the excitation source corresponds to an influence on the field values of the plurality of voxels due to the excitation source during the simulation. 7. The computer-implemented method of claim 5 , further comprising: determining a performance metric of the physical device based on the simulation; determining a loss metric based on a difference between the performance metric and a target performance metric of the physical device; backpropagating the loss metric using the attenuated field values to determine an influence of changes in the structural parameters on the loss metric; and generating a revised description of the physical device by updating the structural parameters to reduce the loss metric. 8. The computer-implemented method of claim 7 , further comprising: determining a field gradient of a first voxel included in the plurality of voxels based on the reverse simulation; determining a loss gradient of the first voxel, wherein the loss gradient is related to changes of the loss metric with respect to the field value of the first voxel over the plurality of time steps; and combining the loss gradient and the field gradient to determine a structural gradient of the first voxel, wherein the structural gradient describes a first influence of changes in the structural value of the first voxel on the loss metric, and wherein the influence of changes in the structural parameters on the loss metric includes the first influence. 9. The computer-implemented method of claim 7 , further comprising: iteratively performing cycles of simulating the physical device, backpropagating the loss metric, and updating the structural parameters to reduce the loss metric until the loss metric substantially converges such that the difference between the performance metric and the target performance metric is within a threshold range. 10. The computer-implemented method of claim 1 , wherein the simulated environment includes one or more perfectly matched layers (“PMLs”) that correspond to outer boundaries of the simulated environment and collectively surround the plurality of voxels included in the simulated environment that describe the physical device, and wherein the PMLs are included in the absorbing boundaries. 11. The computer-implemented method of claim 1 , wherein the physical device includes at least one absorbing medium that attenuates the output of the excitation source during the simulation, and wherein the absorbing medium is included in the absorbing boundaries, wherein the absorbing medium and the one or more absorbing boundaries are not information preserving when performing the simulation, and wherein the attenuation corresponds to the information that is lost during the simulation. 12. The computer-implemented method of claim 11 , wherein the absorbing medium is a non-linear optical material. 13. The computer-implemented method of claim 1 , wherein the attenuated field values are recorded in a compressed representation. 14. The computer-implemented method of claim 1 , wherein the attenuated field values are subsampled during the simulation at less than a one-to-one correspondence with the plurality of time steps such that the attenuated field values are not recorded at one or more of the plurality of time steps. 15. The computer-implemented method of claim 14 , further comprising: interpolating the attenuated field values that have been recorded to generate at least a one-to-one correspondence between the attenuated field values and the plurality of time steps. 16. At least one non-transitory machine-accessible storage medium that provides instructions that, when executed by a machine, will cause the machine to perform operations comprising: receiving an initial description of a physical device represented by structural parameters within a simulated environment; performing a simulation of the physical device in response to an excitation source over a plurality of time steps, wherein the simulated environment includes one or more absorbing boundaries for attenuation of an output of the excitation source during the simulation, wherein the one or more absorbing boundaries surround a plurality of voxels representative of the physical device; recording attenuated field values of the simulated environment associated with the attenuation during the simulation over the plurality of time steps of the simulation; and performing a reverse simulation of the physical device in which the attenuated field values are replayed over the plurality of time steps in reverse to recover a field response of the simulated environment at one or more of the plurality of time steps for one or more voxels included in the plurality of voxels representative of the physical device.
Design optimisation, verification or simulation (optimisation, verification or simulation of circuit designs G06F30/30) · CPC title
Three-dimensional [3D] modelling for computer graphics · CPC title
Mechanical parametric or variational design · CPC title
using finite element methods [FEM] or finite difference methods [FDM] · CPC title
Related publications grouped by family.
Answers are generated from the same data shown on this page.