20240046068. INFORMATION PROCESSING DEVICE FOR IMPROVING QUALITY OF GENERATOR OF GENERATIVE ADVERSARIAL NETWORK (GAN) simplified abstract (NOMURA RESEARCH INSTITUTE, LTD.)

INFORMATION PROCESSING DEVICE FOR IMPROVING QUALITY OF GENERATOR OF GENERATIVE ADVERSARIAL NETWORK (GAN)

Organization Name

NOMURA RESEARCH INSTITUTE, LTD.

Inventor(s)

Teruhiro Tagomori of Irvine CA (US)

INFORMATION PROCESSING DEVICE FOR IMPROVING QUALITY OF GENERATOR OF GENERATIVE ADVERSARIAL NETWORK (GAN) - A simplified explanation of the abstract

This abstract first appeared for US patent application 20240046068 titled 'INFORMATION PROCESSING DEVICE FOR IMPROVING QUALITY OF GENERATOR OF GENERATIVE ADVERSARIAL NETWORK (GAN)

Simplified Explanation

The abstract of the patent application describes an information processing device that uses a conditional generative adversarial network to learn a generative model. This generative model generates data belonging to a designated class based on a noise vector and the designated class. The device also learns a classification model that classifies input data based on whether it belongs to the designated class. A formal verification algorithm is then used to determine whether a property is satisfied when the classification model classifies an output of the generative model. The property indicates that the generative model does not generate data classified into a different class from the designated class, within a certain norm of the noise vector input.

• The information processing device executes a learning algorithm of a conditional generative adversarial network.
• The generative model generates data belonging to a designated class based on a noise vector and the designated class.
• The classification model classifies input data based on whether it belongs to the designated class.
• A formal verification algorithm is used to determine whether a property is satisfied when the classification model classifies an output of the generative model.
• The property ensures that the generative model does not generate data classified into a different class from the designated class, within a certain norm of the noise vector input.

Potential applications of this technology:

• Data generation for specific classes: This technology can be used to generate data belonging to a designated class, which can be useful in various applications such as image synthesis, text generation, or data augmentation for machine learning.
• Conditional data generation: The conditional generative model can generate data based on specific conditions, allowing for targeted data generation for specific scenarios or tasks.

Problems solved by this technology:

• Data quality control: The formal verification algorithm ensures that the generative model does not generate data classified into a different class from the designated class. This helps maintain the quality and integrity of the generated data.
• Class-specific data generation: By learning a generative model that specifically generates data belonging to a designated class, this technology solves the problem of generating data that meets specific class requirements.

Benefits of this technology:

• Improved data generation: The use of a conditional generative adversarial network allows for more precise and targeted data generation, ensuring that the generated data belongs to the designated class.
• Efficient verification: The formal verification algorithm provides a reliable and efficient method to verify the property that the generative model does not generate data classified into a different class. This helps in ensuring the accuracy and reliability of the generated data.

Original Abstract Submitted

an information processing device executes learning a generative model that generates data belonging to a designated class on the basis of a noise vector and the designated class by executing a learning algorithm of a conditional generative adversarial network, learning a classification model that classifies input data based on whether the input data is in the designated class, and determining whether a property is satisfied when the classification model classifies an output of the generative model by executing a formal verification algorithm. the property indicates that the generative model does not generate data classified into a class different from a first class designated for the generative model, within a range of a certain norm of a noise vector input to the generative model.