20230140142. GENERATIVE ADVERSARIAL NEURAL ARCHITECTURE SEARCH simplified abstract (Unknown Organization)

GENERATIVE ADVERSARIAL NEURAL ARCHITECTURE SEARCH

Organization Name

Unknown Organization

Inventor(s)

Seyed Saeed Changiz Rezaei of Vancouver (CA)

Fred Xuefei Han of Edmonton (CA)

Di Niu of Edmonton (CA)

GENERATIVE ADVERSARIAL NEURAL ARCHITECTURE SEARCH - A simplified explanation of the abstract

This abstract first appeared for US patent application 20230140142 titled 'GENERATIVE ADVERSARIAL NEURAL ARCHITECTURE SEARCH

Simplified Explanation

The abstract describes a method and system for neural architectural search (NAS) using a generative adversarial network (GAN). The GAN consists of a generator and a discriminator. The user device sends a query for a neural network architecture, which includes a search space. The generator generates multiple neural network architectures based on the search space. The discriminator then selects the best architecture from the generated options, which is then transmitted back to the user device.

• A generative adversarial network (GAN) is used for neural architectural search (NAS).
• The GAN includes a generator and a discriminator.
• A user device sends a query for a neural network architecture, including a search space.
• The generator generates multiple neural network architectures based on the search space.
• The discriminator selects the optimal architecture from the generated options.
• The optimal architecture is transmitted back to the user device.

Potential Applications

• Automated neural network architecture search.
• Optimization of neural network architectures for specific tasks.
• Efficient exploration of large search spaces for neural architectures.

Problems Solved

• Manual selection of neural network architectures can be time-consuming and inefficient.
• Exploring large search spaces for optimal architectures can be challenging.
• Automating the process of neural architectural search can save time and resources.

Benefits

• Faster and more efficient neural architectural search.
• Improved performance of neural networks for specific tasks.
• Reduction in manual effort required for selecting optimal architectures.

Original Abstract Submitted

a method and system for neural architectural search (nas) for performing a task. a generative adversarial network comprising a generator and a discriminator receives, from a user device, a query for neural network architecture, the query including a search space. the generator of the generative adversarial network generates a plurality of generated neural network architectures responsive to the received search space. the discriminator of the generative adversarial network selects an optimal neural network architecture from among the plurality of generated neural network architectures. the optimal generated neural network architecture is transmitted to the user device.