Email
Password
Remember meForgot password?
    Log in with Twitter

article imageFirst microelectromechanical neural network created

By Tim Sandle     Nov 1, 2018 in Science
Researchers report that a new reservoir computer is the first-ever microelectromechanical neural network application. This could lead to new devices which act simultaneously as both a sensor and a computer.
Scientists have constructed the first reservoir computing device designed with a microelectromechanical system. This is a neural network that utilizes the nonlinear dynamics of a microscale silicon beam to perform complex calculations. Longer-term the researchers aim to develop devices that can function both as a sensor and a computer. This will require only a tiny fraction of the energy required for normal computer use.
Microelectromechanical systems are part of the microscopic devices, particularly those with moving parts, such as nanoelectromechanical systems. Applications of this field include inkjet printers,digital micromirror device displays and accelerometers.
The new research, which comes from the Université de Sherbrooke in Québec, is part of attempts to develop computers whose physical architecture mimics the human brain. According to Phys.org, the device is based on the dynamics of the silicon beam, which oscillates in space. The output from this oscillation can be used to construct a virtual neural network "that projects the input signal into the higher dimensional space required for neural network computing."
Lead researcher Guillaume Dion outlines the computational advantages, when he states: "These kinds of calculations are normally only done in software, and computers can be inefficient. Many of the sensors today are built with MEMS, so devices like ours would be ideal technology to blur the boundary between sensors and computers."
The research has been published in the Journal of Applied Physics. The research paper is titled "Reservoir computing with a single delay-coupled non-linear mechanical oscillator."
More about neural networks, microelectromechanical, Network, Data
More news from
Latest News
Top News