In this article:

Panasonic had 33 patents in artificial intelligence during Q3 2023. Panasonic Holdings Corp filed patents related to various technologies in Q3 2023. One patent focuses on a training method for a neural network model that involves supervised contrastive learning for representation learning and computer vision tasks. Another patent relates to generating a map of a vehicle’s passenger compartment and controlling vehicle features based on the map. Additionally, there are patents for optimizing parameters and hyperparameters in deep neural networks and a control device for wireless communication that adapts to changing environments. Lastly, there is a patent for automated machine learning supply chain planning using a digital image representation and an auto-encoder model. GlobalData’s report on Panasonic gives a 360-degreee view of the company including its patenting strategy. Buy the report here.

Smarter leaders trust GlobalData

Report-cover

Premium Insights Panasonic Holdings Corp - Company Profile

Buy the Report

Premium Insights

The gold standard of business intelligence.

Find out more

Panasonic grant share with artificial intelligence as a theme is 52% in Q3 2023. Grant share is based on the ratio of number of grants to total number of patents.

Recent Patents

Application: Training method (Patent ID: US20230274533A1)

The patent filed by Panasonic Holdings Corp. describes a training method for a neural network model that includes two network branches: one for representation learning and the other for computer vision tasks. The method involves obtaining image data and labels, performing data augmentation, extracting feature representations, and training the encoder network model and the first model using a first loss function. The method also includes inferring labels using the second model and training the encoder network model and the second model using a second loss function. The training of both models is performed simultaneously.

The training method includes several steps. First, N image data items and their corresponding labels are obtained from a dataset. Data augmentation is then performed to obtain M image data items and labels. The encoder network model extracts feature representations from the M image data items, and the first model projects these representations onto embedding vectors for supervised contrastive learning. Label processing is performed to convert the labels of the M image data items into labels applicable to the computer vision tasks. The encoder network model and the first model are trained using a first loss function, the labels of the M image data items, and the embedding vectors.

Next, the M image data items resulting from data augmentation are obtained, and feature representations are extracted from them using the encoder network model. The second model infers labels for the M image data items based on these feature representations. The encoder network model and the second model are then trained using a second loss function, the inferred labels, and the labels of the M image data items. The training of both models is performed simultaneously.

The patent also includes additional claims. Claim 2 describes the label processing step, which involves converting the labels into applicable representations with values of 0 or 1 for each dimension. Claim 3 specifies that when the M image data items include two different items obtained from the same image data item, supervised contrastive loss is calculated using the first loss function. Claim 4 states that if the applicable representations have a value of 1 for two or more dimensions, these values are further converted to 0. Claim 5 mentions that a loss based on vector similarity is calculated during the training of the encoder network model and the first model. Claim 6 introduces a third model in the first network branch, which predicts a second embedding vector from a first embedding vector. Claim 7 states that a loss based on cosine similarity is calculated during the training of the encoder network model, the first model, and the third model.

To know more about GlobalData’s detailed insights on Panasonic, buy the report here.

Premium Insights

From

The gold standard of business intelligence.

Blending expert knowledge with cutting-edge technology, GlobalData’s unrivalled proprietary data will enable you to decode what’s happening in your market. You can make better informed decisions and gain a future-proof advantage over your competitors.

GlobalData

GlobalData, the leading provider of industry intelligence, provided the underlying data, research, and analysis used to produce this article.

GlobalData Patent Analytics tracks bibliographic data, legal events data, point in time patent ownerships, and backward and forward citations from global patenting offices. Textual analysis and official patent classifications are used to group patents into key thematic areas and link them to specific companies across the world’s largest industries.