The team develops a method for calculating neural networks in water

The team develops a method for calculating neural networks in water

Computer neural network in water

An ion circuit comprising hundreds of ion transistors. Credit: Woo-Bin Jung/Harvard SEAS

Microprocessors in smartphones, computers, and data centers process information by manipulating electrons through solid semiconductors, but our brains have a different system. They rely on the manipulation of ions in a liquid to process information.

Inspired by the brain, researchers have long sought to develop “ionics” in aqueous solution. While ions in water move more slowly than electrons in semiconductors, scientists believe that the diversity of ion species with different physical and chemical properties could be harnessed for richer and more efficient information processing. diversified.

Ion computing, however, is still in its infancy. To date, laboratories have only developed individual ionic devices such as ionic diodes and transistors, but no one has so far assembled many such devices into a more complex computer circuit.

A team of researchers from the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS), in collaboration with DNA Script, a biotech startup, developed an ion circuit comprising hundreds of ion transistors and carried out a basic calculation process of neural network.

The research is published in Advanced materials.

The researchers started by building a new type of ion transistor using a technique they recently developed. The transistor consists of an aqueous solution of quinone molecules, interfaced with two concentric annular electrodes with a central electrode in the shape of a disc, like a bull’s eye. The two ring electrodes electrochemically lower and adjust the local pH around the central disc by producing and trapping hydrogen ions. A voltage applied to the central disc causes an electrochemical reaction to generate an ion current from the disc into the water. The rate of reaction can be accelerated or decreased – by increasing or decreasing the ion current – by adjusting the local pH. In other words, the pH controls (gates) the ionic current of the disk in the aqueous solution, creating an ionic counterpart of the electronic transistor.

Computer neural network in water

A CMOS chip (left) with an array (center) of hundreds of individual ion transistors (right). Credit: Woo-Bin Jung/Harvard SEAS

They then designed the pH-triggered ion transistor so that the disk current was an arithmetic multiplication of the disk voltage and a “weight” parameter representing the local pH triggering the transistor. They organized these transistors into a 16×16 network to extend analog arithmetic multiplication of individual transistors into analog matrix multiplication, with the network of local pH values ​​serving as the weighting matrix found in neural networks.

“Matrix multiplication is the most popular computation in neural networks for artificial intelligence,” said Woo-Bin Jung, postdoctoral fellow at SEAS and first author of the paper. “Our ion circuit performs matrix multiplication in water in an analog fashion based entirely on electrochemical machinery.”

“Microprocessors manipulate electrons digitally to perform matrix multiplication,” said Donhee Ham, Gordon McKay Professor of Electrical Engineering and Applied Physics at SEAS and lead author of the paper. “Although our ion circuit cannot be as fast or accurate as digital microprocessors, the multiplication of the electrochemical matrix in water is charming in itself and has the potential to be energy efficient.”

Now the team is looking to enrich the chemical complexity of the system.

“So far, we have used only 3-4 ion species, such as hydrogen and quinone ions, to enable ion gating and transport in the aqueous ion transistor,” Jung said. “It will be very interesting to use more diversified ionic species and to see how we can exploit them to enrich the content of the information to be processed.”

The research was co-authored by Han Sae Jung, Jun Wang, Henry Hinton, Maxime Fournier, Adrian Horgan, Xavier Godron, and Robert Nicol.


Chemists suggest using polymer ionic liquids in supercapacitors


More information:
Woo-Bin Jung et al, A watery analog MAC machine, Advanced materials (2022). DOI: 10.1002/adma.202205096

Provided by Harvard John A. Paulson School of Engineering and Applied Sciences

Quote: Team develops method for neural net computing in water (2022, September 29) retrieved September 29, 2022 from https://phys.org/news/2022-09-team-method-neural-net.html

This document is subject to copyright. Except for fair use for purposes of private study or research, no part may be reproduced without written permission. The content is provided for information only.


#team #develops #method #calculating #neural #networks #water

Leave a Comment

Your email address will not be published. Required fields are marked *