公开(公告)号：US12211166B2

公开(公告)日：2025-01-28

申请号：US18386515

申请日：2023-11-02

Applicant: Snap Inc.

Inventor： Gleb Dmukhin ,  Egor Nemchinov ,  Yurii Volkov

IPC: G06T19/20 ,  G06N3/02 ,  G06N3/047 ,  G06N3/0475 ,  G06N3/0895 ,  G06N3/094 ,  G06T17/20 ,  H04L51/10

Abstract: A messaging system processes three-dimensional (3D) models to generate ground truths for training machine learning models for applications of the messaging system. A method of generating ground truths for machine learning includes generating a plurality of first rendered images from a first 3D base model where each first rendered image includes the 3D base model modified by first augmentations of a plurality of augmentations. The method further includes determining for a second 3D base model incompatible augmentations of the first plurality of augmentations, where the incompatible augmentations indicate changes to fixed features of the second 3D base model, and generating a plurality of second rendered images from a second 3D base model, each second rendered image comprising the second 3D base model modified by second augmentations, the second augmentations corresponding to the first augmentations of a corresponding first rendered image, where the second augmentations comprises augmentations of the first augmentations that are not incompatible augmentations.

公开(公告)号：US20250029372A1

公开(公告)日：2025-01-23

申请号：US18776010

申请日：2024-07-17

Applicant: Arizona Board of Regents on behalf of Arizona State University

Inventor： Zuwei Guo ,  Nahid Ul Islam ,  Jianming Liang

IPC: G06V10/774 ,  G06N3/0455 ,  G06N3/088 ,  G06N3/094 ,  G06V10/82 ,  G16H30/40

Abstract: A stepwise incremental pre-training for integrating discriminative, restorative, and adversarial learning in an AI model. Exemplary systems include means for receiving a training dataset for training a unified AI model. The unified AI model applies each of discriminative, restorative, and adversarial learning operations through three transferable components: a discriminative encoder, a restorative decoder, and an adversarial encoder. Stepwise incremental pre-training operations train the unified AI model, including pre-training the discriminative encoder via discriminative learning and attaching the trained discriminative encoder with the restorative decoder to form a skip-connected encoder-decoder, pre-training the skip-connected encoder-decoder via joint discriminative and restorative learning, and associating the pre-trained skip-connected encoder-decoder with the adversarial encoder. Training of the AI model is finalized by performing full discriminative, restorative, and adversarial learning on the training dataset using the unified AI model.

公开(公告)号：US12182720B2

公开(公告)日：2024-12-31

申请号：US17044399

申请日：2018-06-29

Applicant: NEC Corporation

Inventor： Shivangi Mahto

IPC: G06N3/0475 ,  G06F18/10 ,  G06F18/21 ,  G06F18/2132 ,  G06F18/2411 ,  G06F18/2415 ,  G06N3/045 ,  G06N3/047 ,  G06N3/088 ,  G06N3/094

Abstract: An apparatus for pattern recognition includes a generator which transforms noisy feature vectors into denoised feature vectors, a discriminator which takes the denoised feature vectors and the original clean feature vectors corresponding to the denoised feature vectors as input and predicts probability for both of the input features of being an original clean feature, classifies the input feature vectors into its corresponding classes, an objective function calculator which calculates generator and discriminator losses using the denoised feature vectors, the clean feature vectors from which the noisy feature vectors have been made, the estimated classes and their true classes, and a Parameter updater which updates parameters of the generator and the discriminator according to loss minimization.

公开(公告)号：US20240412437A1

公开(公告)日：2024-12-12

申请号：US18332300

申请日：2023-06-09

Applicant: Mercedes-Benz Group AG

Inventor： Thomas Albin ,  Timothy Fenlon

IPC: G06T13/20 ,  G06N3/094

Abstract: Methods, computing systems, and technology for generative modeling of wheel hub display content are presented. A control circuit can: obtain user input data including a description of content to be presented via a display device positioned on a wheel of the vehicle; generate, using one or more models including a machine-learned generative model, the content based on the user input data; receive an output of the one or more models, the output including the generated content; and provide, for presentation via the display device positioned on the wheel of the vehicle, data indicative of the generated content. The machine-learned generative model can be trained to process the user input data and provide generated content that is: (i) based on the description of the content included in the user input data, and (ii) configured for presentation via the display device positioned on the wheel of the vehicle.

公开(公告)号：US20240377343A1

公开(公告)日：2024-11-14

申请号：US18684558

申请日：2022-08-22

Applicant: ASML NETHERLANDS B.V.

Inventor： Yunbo GUO ,  Zhu WANG ,  Feng YANG ,  Qian ZHAO ,  Mu FENG ,  Jen-Shiang WANG

IPC: G01N23/2251 ,  G06N3/045 ,  G06N3/0475 ,  G06N3/094

Abstract: Described herein is a metrology system and a method for converting metrology data via a trained machine learning (ML) model. The method includes accessing a first (MD1) SEM data set (e.g., images, contours, etc.) acquired by a first scanning electron metrology (SEM) system (TS1) and a second (MD2) SEM data set acquired by a second SEM system (TS2), where the first SEM data set and the second SEM data set being associated with a patterned substrate. Using the first SEM data set and the second SEM data set as training data, a machine learning (ML) model is trained (P303) such that the trained ML model is configured to convert (P307) a metrology data set (310) acquired (P305) by the second SEM system to a converted data set (311) having characteristics comparable to metrology data being acquired by the first SEM system. Furthermore, measurements may be determined based on the converted SEM data.

公开(公告)号：US12122417B2

公开(公告)日：2024-10-22

申请号：US17884384

申请日：2022-08-09

Applicant: Motional AD LLC

Inventor： Edouard Francois Marc Capellier

IPC: B60W60/00 ,  G06N3/0475 ,  G06N3/094

CPC classification number: B60W60/001 ,  G06N3/0475 ,  G06N3/094 ,  B60W2556/35

Abstract: Provided are methods for detecting when a vehicle is encountering an out of operational design domain (ODD) scenario, which can include training a generative adversarial network (GAN) including a generator network and a discriminator network. The generator network may be trained to generate synthesized scenarios. The discriminator network may be trained to distinguish between true scenarios and the synthesized scenarios generated by the generator network. The trained discriminator network may be applied to detect when a vehicle encounters an out of operational design domain (ODD) scenario. Some methods described also include controlling the motion of the vehicle in response to an output of the trained discriminator network indicating that the vehicle is encountering the out of operational design domain (ODD) scenario. Systems and computer program products are also provided.

公开(公告)号：US12118472B2

公开(公告)日：2024-10-15

申请号：US18521987

申请日：2023-11-28

Applicant: Treble Technologies

Inventor： Finnur Pind ,  Jesper Pedersen

IPC: G06N3/094 ,  G06N3/0455 ,  H04S7/00

CPC classification number: G06N3/094 ,  G06N3/0455 ,  H04S7/302 ,  H04S7/303 ,  H04S7/305 ,  H04S7/307 ,  H04S2400/11 ,  H04S2400/15

Abstract: A computer implemented method for training a machine learning model for audio compensation includes: receiving 3D models of a plurality of rooms, each of the 3D models comprising at least one sound source and at least one acoustic property; receiving a plurality of impulse responses at a listening position in each of the plurality of rooms, training the machine learning model for audio compensation using at least the plurality of impulse responses as input. Methods and systems using the disclosed computer implemented method and its features are also disclosed.

公开(公告)号：US20240303502A1

公开(公告)日：2024-09-12

申请号：US18281497

申请日：2022-03-10

Applicant: Google LLC

Inventor： Yuezhen Niu ,  Hartmut Neven ,  Vadim Smelyanskiy ,  Sergio Boixo Castrillo

IPC: G06N3/094 ,  G06N10/60

CPC classification number: G06N3/094 ,  G06N10/60

Abstract: Methods and apparatus for learning a target quantum state. In one aspect, a method for training a quantum generative adversarial network (QGAN) to learn a target quantum state includes iteratively adjusting parameters of the QGAN until a value of a QGAN loss function converges, wherein each iteration comprises: performing an entangling operation on a discriminator network input of a discriminator network in the QGAN to measure a fidelity of the discriminator network input, wherein the discriminator network input comprises the target quantum state and a first quantum state output from a generator network in the QGAN, wherein the first quantum state approximates the target quantum state; and performing a minimax optimization of the QGAN loss function to update the QGAN parameters, wherein the QGAN loss function is dependent on the measured fidelity of the discriminator network input.

公开(公告)号：US20240289921A1

公开(公告)日：2024-08-29

申请号：US18130022

申请日：2023-04-03

Applicant: Dell Products L.P.

Inventor： Zijia Wang ,  Zhisong Liu ,  Zhen Jia

IPC: G06T3/40 ,  G06N3/0464 ,  G06N3/094

CPC classification number: G06T3/4046 ,  G06N3/0464 ,  G06N3/094

Abstract: Embodiments of the present disclosure relate to a method, an electronic device, and a computer program product for training an image processing model. The method in an illustrative embodiment includes: obtaining a folding weight of a folded convolutional layer of a pre-trained generator by performing a folding operation on a plurality of weights of a plurality of convolutional layers of the pre-trained generator. The method further includes: embedding the pre-trained generator into the image processing model. The method further includes: training the image processing model using a plurality of pairs of sample images, wherein at least one pair of sample images of the plurality of pairs of sample images includes a first sample image having a first resolution and a second sample image having a second resolution, and wherein the first resolution is less than the second resolution.

公开(公告)号：US20240256889A1

公开(公告)日：2024-08-01

申请号：US18565510

申请日：2021-05-31

Applicant: ROBERT BOSCH GMBH ,  TSINGHUA UNIVERSITY

Inventor： Hang Su ,  Jun Zhu ,  Tianyu Pang ,  Xiao Yang ,  Yinpeng Dong ,  Zhijie Deng ,  Ze Cheng

IPC: G06N3/094

CPC classification number: G06N3/094

Abstract: A method for deep learning. The method includes: receiving, by a deep learning model, a plurality of samples and a plurality of labels corresponding to the plurality of samples; adversarially augmenting, by the deep learning model, the plurality of samples based on a threat model; and assigning, by the deep learning model, a low predictive confidence to one or more adversarially augmented samples of the plurality of adversarially augmented samples having noisy labels due to the adversarially augmenting based on the threat model.