MIT News August 22, 2023 Optical neural networks (ONNs) have recently emerged to process deep neural networks (DNN) tasks with high clock rates, parallelism, and low-loss data transmission. However, existing challenges for ONNs are high energy consumption due to their low electro-optic conversion efficiency, low compute density due to large device footprints and channel crosstalk, and long latency due to the lack of inline nonlinearity. An international team of researchers (USA – MIT, UCLA, industry, Germany) experimentally demonstrated a spatial-temporal-multiplexed ONN system that simultaneously overcomes all these challenges. They exploited neuron encoding with volume-manufactured micrometre-scale vertical-cavity surface-emitting laser (VCSEL) arrays […]
Tag Archives: Neural networks
Navigating the future of underwater geolocalization: How polarization patterns enable new technology
Science Daily July 10, 2023 Current methods for underwater geolocalization rely on tethered systems with limited coverage or daytime imagery data in clear waters, leaving much of the underwater environment unexplored. Geolocalization in turbid waters or at night has been considered unfeasible due to absence of identifiable landmarks. Researchers at the University of Illinois, Urbana) have developed a novel method for underwater geolocalization using deep neural networks trained on ∼10 million polarization-sensitive images acquired globally, along with camera position sensor data. They achieved longitudinal accuracy of ∼55 km (∼1000 km) during daytime (nighttime) at depths up to ∼8 m, regardless […]
Drones navigate unseen environments with liquid neural networks
MIT News April 19, 2023 Autonomous robots can learn to perform visual navigation tasks from offline human demonstrations and generalize online and unseen scenarios within the same environment they have been trained on. It is challenging for these agents to take a step further and robustly generalize to new environments with drastic scenery changes that they have never encountered. Researchers at MIT have developed a method to create robust flight navigation agents that successfully perform vision-based fly-to-target tasks beyond their training environment under drastic distribution shifts. They designed an imitation learning framework using liquid neural networks, a brain-inspired class of […]
Neural networks could help predict destructive earthquakes
Phys.org March 3, 2023 The movement and deformation of the Earth’s crust and upper mantle provide critical insights into the evolution of earthquake processes and future earthquake potentials. Crustal deformation can be modeled by dislocation models that represent earthquake faults in the crust as defects in a continuum medium. Researchers in Japan have proposed a physics-informed deep learning approach to model crustal deformation due to earthquakes. Neural networks can represent continuous displacement fields in arbitrary geometrical structures and mechanical properties of rocks by incorporating governing equations and boundary conditions into a loss function. They introduced polar coordinate system to accurately […]
Neural networks predict forces in jammed granular solids
Phys.org September 1, 2022 Force chains are quasi-linear self-organised structures carrying large stresses and are ubiquitous in jammed amorphous materials like granular materials, foams or even cell assemblies. Understanding force chains is crucial in describing the mechanical and transport properties of granular solids and this applies in a wide range of circumstances—for example how sound propagates or how sand responds to mechanical deformation. Predicting where they will form upon deformation is crucial to describe the properties of such materials but remains an open question. An international team of researchers (Germany, Belgium, UK) demonstrated that graph neural networks (GNN) can accurately […]
Neural networks and ‘ghost’ electrons accurately reconstruct behavior of quantum systems
Phys.org August 3, 2022 Predicting the properties of a molecule or material requires calculating the collective behavior of its electrons because the electrons can become “quantum mechanically” entangled with one another. The entangled web of connections becomes tricky for even the most powerful computers to unravel directly for any system with more than a handful of particles. An international team of researchers (USA – Res. org., Switzerland) created a way to simulate entanglement by adding to their computations extra “ghost” electrons that interact with the system’s actual electrons. The behavior of the added electrons is controlled by neural network. The […]
The future of data storage is double-helical, research indicates
Science Daily March 3, 2022 A team of researchers in the US (University of Illinois, UMass Amherst, Stanford University) expanded molecular alphabet for DNA data storage comprising four natural and seven chemically modified nucleotides that are readily detected and distinguished using nanopore sequencers. They showed that Mycobacterium smegmatis porin A (MspA) nanopores can accurately discriminate 77 combinations and orderings of chemically diverse monomers within homo- and heterotetrameric sequences. The sequencing accuracy exceeded 60%. The extended molecular alphabet may potentially offer a nearly 2-fold increase in storage density and potentially the same order of reduction in the recording latency, thereby enabling […]
Pushing computing to the edge by rethinking microchips’ design
EurekAlert February 24, 2021 Two years ago, researchers at Princeton University fabricated a new chip designed to improve the performance of neural networks. The chip performed tens to hundreds of times better than other advanced microchips. But the chip’s major drawback was that it uses a very unusual and disruptive architecture as it needs to be reconciled with the massive amount of infrastructure and design methodology that we have and use today. Now the team has created software that would allow the new chips to work with different types of networks, allow the systems to be scalable both in hardware […]