UCLA engineers have made major improvements on their design of an optical neural network -a device inspired by how the human brain works - that can identify objects or process information at the speed of light.
The development could lead to intelligent camera systems that figure out what they are seeing simply by the patterns of light that run through a 3D engineered material structure. Their new design takes advantage of the parallelization and scalability of optical-based computational systems.
For example, such systems could be incorporated into self-driving cars or robots, helping them make near-instantaneous decisions faster and using less power than computer-based systems that need additional time to identify an object after it's been seen.
The technology was first introduced by the UCLA group in 2018. The system uses a series of 3D-printed wafers or layers with uneven surfaces that transmit or reflect incoming light - they're reminiscent in look and effect to frosted glass.
These layers have tens of thousands of pixel points - essentially these are artificial neurons that form an engineered volume of material that computes all-optically. Each object will have a unique light pathway through the 3D fabricated layers.
Behind those layers are several light detectors, each previously assigned in a computer to deduce what the input object is by where the most light ends up after traveling through the layers.
For example, if it's trained to figure out handwritten digits, then the detector programmed to identify a "5" will see the most of the light hit that detector after the image of a "5" has traveled through the layers.
In this recent study, published in the open access journal Advanced Photonics, the UCLA researchers have significantly increased the system's accuracy by adding a second set of detectors to the system, and therefore each object type is now represented with two detectors rather than one.
The researchers aimed to increase the signal difference between a detector pair assigned to an object type. Intuitively, this is similar to weighing two stones simultaneously with the left and right hands - it is easier this way to differentiate if they are of similar weight or have different weights.
This differential detection scheme helped UCLA researchers improve their prediction accuracy for unknown objects that were seen by their optical neural network.
"Such a system performs machine-learning tasks with light-matter interaction and optical diffraction inside a 3D fabricated material structure, at the speed of light and without the need for extensive power, except the illumination light and a simple detector circuitry," said Aydogan Ozcan, Chancellor's Professor of Electrical and Computer Engineering and the principal investigator on the research.
"This advance could enable task-specific smart cameras that perform computation on a scene using only photons and light-matter interaction, making it extremely fast and power efficient."
The researchers tested their system's accuracy using image datasets of hand-written digits, items of clothing, and a broader set of various vehicles and animals known as the CIFAR-10 image dataset. They found image recognition accuracy rates of 98.6%, 91.1% and 51.4% respectively.
Those results compare very favorably to earlier generations of all-electronic deep neural nets. While more recent electronic systems have better performance, the researchers suggest that all-optical systems have advantages in inference speed, low-power, and can be scaled up to accommodate and identify many more objects in parallel.
Other authors on the study include graduate students Jingxi Li, Deniz Mengu and Yi Luo; and Yair Rivenson, a UCLA assistant adjunct professor of electrical and computer engineering.
Ozcan also has UCLA faculty appointments in bioengineering and in surgery at the David Geffen School of Medicine. He is the associate director of the UCLA California NanoSystems Institute and is an HHMI professor.
The study was supported by the Koç Group, the National Science Foundation and the Howard Hughes Medical Institute.
Amy Akmal | EurekAlert!
Quantum computers by AQT and University of Innsbruck leverage Cirq for quantum algorithm development
16.09.2019 | Universität Innsbruck
Artificial Intelligence speeds up photodynamics simulations
12.09.2019 | University of Vienna
Later during this century, around 2060, a paradigm shift in global energy consumption is expected: we will spend more energy for cooling than for heating....
Researchers from the Department of Atomically Resolved Dynamics of the Max Planck Institute for the Structure and Dynamics of Matter (MPSD) at the Center for Free-Electron Laser Science in Hamburg, the University of Potsdam (both in Germany) and the University of Toronto (Canada) have pieced together a detailed time-lapse movie revealing all the major steps during the catalytic cycle of an enzyme. Surprisingly, the communication between the protein units is accomplished via a water-network akin to a string telephone. This communication is aligned with a ‘breathing’ motion, that is the expansion and contraction of the protein.
This time-lapse sequence of structures reveals dynamic motions as a fundamental element in the molecular foundations of biology.
Two research teams have succeeded simultaneously in measuring the long-sought Thorium nuclear transition, which enables extremely precise nuclear clocks. TU Wien (Vienna) is part of both teams.
If you want to build the most accurate clock in the world, you need something that "ticks" very fast and extremely precise. In an atomic clock, electrons are...
Researchers from Chalmers University of Technology have demonstrated a detector made from graphene that could revolutionize the sensors used in next-generation space telescopes. The findings were recently published in the scientific journal Nature Astronomy.
Beyond superconductors, there are few materials that can fulfill the requirements needed for making ultra-sensitive and fast terahertz (THz) detectors for...
A supersolid is a state of matter that can be described in simplified terms as being solid and liquid at the same time. In recent years, extensive efforts have been devoted to the detection of this exotic quantum matter. A research team led by Tilman Pfau and Tim Langen at the 5th Institute of Physics of the University of Stuttgart has succeeded in proving experimentally that the long-sought supersolid state of matter exists. The researchers report their results in Nature magazine.
In our everyday lives, we are familiar with matter existing in three different states: solid, liquid, or gas. However, if matter is cooled down to extremely...
10.09.2019 | Event News
04.09.2019 | Event News
29.08.2019 | Event News
16.09.2019 | Life Sciences
16.09.2019 | Materials Sciences
16.09.2019 | Health and Medicine