View The Document

Accession Number:

AD1192952

Title:

Computing with Dynamical Systems: Reservoir Computers

Author(s):

• Carroll, Thomas L

Author Organization(s):

• NAVAL RESEARCH LAB WASHINGTON DC

Report Date:

2023-02-09

Abstract:

A reservoir computer is a type of neuromorphic computer that is easy to train. Reservoir computers may be build as highly parallel analog systems, making them potentially useful for edge computing applications, such as unmanned arial vehicles. Obtaining optimal performance from a reservoir computer requires computationally expensive optimization procedures. This work unit applied nonlinear dynamics principles to reservoir computers to establish principles for designing optimal reservoir computers that require less computation than blind optimization.

Document Type:

Conference:

Journal:

Pages:

32

File Size:

2.29MB

Contracts:

Grants:

Descriptors:

• artificial intelligence
• neural networks
• machine learning
• nonlinear dynamics
• standards
• information science
• eigenvalues
• equations
• recurrent neural networks
• statistics
• variational equations
• computers
• data science
• differential equations
• network science
• sine waves
• covariance

Identifiers:

• Machine Learning; Artifical intelligence; Neurimorphic; Dynamical systems

Communities of Interest:

• Materials and Manufacturing Processes

Distribution Statement:

Approved For Public Release

View The Document