US Patent Application 18349089. NEURAL NETWORKS WITH SWITCH LAYERS simplified abstract

NEURAL NETWORKS WITH SWITCH LAYERS

Organization Name

Google LLC

Inventor(s)

William Bradley Fedus of Menlo Park CA (US)

Barret Zoph of San Francisco CA (US)

Noam M. Shazeer of Palo Alto CA (US)

NEURAL NETWORKS WITH SWITCH LAYERS - A simplified explanation of the abstract

This abstract first appeared for US patent application 18349089 titled 'NEURAL NETWORKS WITH SWITCH LAYERS

Simplified Explanation

The patent application describes methods, systems, and apparatus for using machine learning to process network inputs and generate network outputs.

• The invention involves a neural network that is designed to perform the machine learning task.
• The neural network includes one or more switch layers, which are specific layers within the network architecture.
• The purpose of these switch layers is not explicitly mentioned in the abstract, but they likely serve a specific function in enhancing the machine learning capabilities of the network.
• The patent application also mentions computer programs encoded on a computer storage medium, indicating that the invention may involve software implementation.
• The abstract does not provide specific details about the machine learning task or the type of network input and output, leaving those aspects open to interpretation.

Original Abstract Submitted

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for performing a machine learning task on a network input to generate a network output. In one aspect, one of the systems includes a neural network configured to perform the machine learning task, the neural network including one or more switch layers.