Mitsubishi Electric had 175 patents in artificial intelligence during Q4 2023. Mitsubishi Electric Corp’s patents in Q4 2023 include a system for automated construction of stochastic deep neural network architectures, a communication satellite system enabling communication between satellites in different orbits, an occupant state adjustment device based on estimation results, an object detection device using multiple feature maps for improved detection, and an awakening level estimation device utilizing occupant state information and machine learning for estimating awakening levels accurately. GlobalData’s report on Mitsubishi Electric gives a 360-degreee view of the company including its patenting strategy. Buy the report here.
Smarter leaders trust GlobalData
Mitsubishi Electric grant share with artificial intelligence as a theme is 34% in Q4 2023. Grant share is based on the ratio of number of grants to total number of patents.
Recent Patents
Application: Automated variational inference using stochastic models with irregular beliefs (Patent ID: US20230419075A1)
The patent filed by Mitsubishi Electric Corp. describes a system and method for automated construction of a stochastic deep neural network (DNN) architecture. The invention focuses on searching for relevant stochastic modes underlying datasets for variational Bayesian inference, allowing for the use of heterogeneous, irregular, and mismatched beliefs in stochastic sampling for intermediate representation in DNNs. The system enables an automatically tuning mechanism of posterior, prior, and likelihood models to enhance generative models and uncertainty models for machine learning tasks. By incorporating adjustable discrepancy measures to regularize intermediate representation using divergence metrics like Renyi's alpha, beta, and gamma divergences, the invention facilitates diverse mixture combinations of stochastic models for probabilistic relations, improving the representation capability of various stochastic DNNs.
The system includes a stochastic DNN block that identifies task labels from multi-dimensional signals through stochastic nodes specified by irregular beliefs for posterior, prior, and likelihood distributions. The system utilizes a memory bank to store datasets, irregular beliefs, hyperparameters, and trainable parameters, with a processor executing probabilistic inference using importance-weighted accumulation after variational sampling at stochastic nodes. The stochastic DNN block is configured with various layers, interconnections, activations, and regularization layers, while the variational sampling employs a reparameterization trick based on variational parameters specified by irregular beliefs. Additionally, the system incorporates hyperparameters for training, architecture, and deployment, along with exploring different values for irregular beliefs and hyperparameters using a hypergradient method for enhanced data analysis.
To know more about GlobalData’s detailed insights on Mitsubishi Electric, buy the report here.
Data 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, 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.
