Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog – Page 8

The current conventional wisdom on deep neural networks (DNNs) is that, in most cases, simply scaling up a model’s parameters and adopting computationally intensive architectures will result in large performance improvements. Although this scaling strategy has proven successful in research labs, real-world industrial deployments introduce a number of complications, as developers often need to repeatedly train a DNN, transmit it to different devices, and ensure it can perform under various hardware constraints with minimal accuracy loss.

The research community has thus become increasingly interested in reducing such models’ storage size on devices while also improving their run-time. Explorations in this area have tended to follow one of two avenues: reducing model size via compression techniques, or using model pruning to reduce computation burdens.

In the new paper LilNetX: Lightweight Networks with EXtreme Model Compression and Structured Sparsification, a team from the University of Maryland and Google Research proposes a way to “bridge the gap” between the two approaches with LilNetX, an end-to-end trainable technique for neural networks that jointly optimizes model parameters for accuracy, model size on the disk, and computation on any given task.

Read more | >

A team of astronomers led by Michael Janssen (Radboud University, The Netherlands) has trained a neural network with millions of synthetic black hole data sets. Based on the network and data from the Event Horizon Telescope, they now predict, among other things, that the black hole at the center of our Milky Way is spinning at near top speed.

The astronomers have published their results and methodology in three papers in the journal Astronomy & Astrophysics.

In 2019, the Event Horizon Telescope Collaboration released the first image of a supermassive black hole at the center of the galaxy M87. In 2022, they presented an image of the black hole in our Milky Way, Sagittarius A*. However, the data behind the images still contained a wealth of hard-to-crack information. An international team of researchers trained a neural network to extract as much information as possible from the data.

Read more | >

Open-source deep-learning framework for building, training, and fine-tuning deep learning models using state-of-the-art Physics-ML methods — NVIDIA/physicsnemo

Read more | >

The National Institute of Information and Communications Technology of Japan, in collaboration with Sony Semiconductor Solutions Corporation (Sony), has developed the world’s first practical surface-emitting laser that employs quantum dot (QD) as the optical gain medium for use in optical fiber communication systems.

This achievement was made possible by NICT’s high-precision technology and Sony’s advanced semiconductor processing technology. The surface-emitting laser developed in this study incorporates nanoscale semiconductor structures called as light-emitting materials. This innovation not only facilitates the miniaturization and reduced power consumption of light sources in optical fiber communications systems but also offers potential cost reductions through and enhanced output via integration.

The results of this research are published in Optics Express.

Read more | >

A team of researchers at the Facility for Rare Isotope Beams (FRIB) at Michigan State University (MSU) has discovered that cobalt-70 isotopes form different nuclear shapes when their energy levels differ only slightly. The findings, published in Nature Communications Physics, shed light on the dynamic, complex nature of exotic nuclear particles.

The team included Artemis Spyrou, professor of physics at the Facility for Rare Isotope Beams (FRIB) and in the MSU Department of Physics and Astronomy, Sean Liddick, associate professor of chemistry at FRIB and in the MSU Department of Chemistry and Experimental Nuclear Science Department head at FRIB, Alex Brown, professor of physics at FRIB, and Cade Dembski, former FRIB student research assistant. Dembski, now working on his Ph.D. at the University of Notre Dame, served as the paper’s lead author.

“When we first started this project, it was motivated by the astrophysical side of nuclear science research, instead of focusing on ,” Dembski said. “As we continued with our , though, we couldn’t quite understand all of the patterns we were seeing. It turned out the reason was due to some interesting nuclear structure effects that we were not expecting, and we ended up writing the paper about those effects.”

Read more | >

Astronomers have conducted very long baseline Interferometry (VLBI) observations of an active galaxy known as Markarian 110. As a result, they detected a relativistic jet in this galaxy. The finding was reported in a research paper published June 4 on the arXiv pre-print server.

Active galactic nuclei (AGNs) are small regions at the center of an active galaxy dominated by the light emitted by dust and gas. Narrow-line Seyfert 1 (NLS1) galaxies are a class of AGNs exhibiting excessive behavior at all wavelengths. They show peculiar characteristics like narrow Balmer lines, strong ionized iron emission lines, and extreme properties in the X-rays.

Markarian 110 (or Mrk 110 for short) is a radio-quiet AGN and an NLS1 at a redshift of 0.035. The galaxy has an apparent magnitude of 15.4 mag and showcases a highly irregular morphology, which suggests a recent interaction or a merging event in this system. It also has a variable core confined to an extremely compact region.

Read more | >

You may not have heard of tantalum, but chances are you’re holding some right now. It’s an essential component in our cell phones and laptops, and currently, there’s no effective substitute. Even if you plan to recycle your devices after they die, the tantalum inside is likely to end up in a landfill or shipped overseas, being lost forever.

As a researcher focused on critical materials recovery, I’ve spent years digging through , not seeing it as garbage, but as an urban mine filled with valuable materials like .

Read more | >

String theory has long been touted as physicists’ best candidate for describing the fundamental nature of the universe, with elementary particles and forces described as vibrations of tiny threads of energy. But in the early 21st century, it was realized that most of the versions of reality described by string theory’s equations cannot match up with observations of our own universe.

In particular, conventional ’s predictions are incompatible with the observation of dark energy, which appears to be causing our universe’s expansion to speed up, and with viable theories of quantum gravity, instead predicting a vast ‘swampland’ of impossible universes.

Now, a new analysis by FQxI physicist Eduardo Guendelman, of Ben-Gurion University of the Negev, in Israel, shows that an exotic subset of string models—in which the of strings is generated dynamically—could provide an escape route out of the string theory swampland.

Read more | >

A team of researchers at the University of California, Los Angeles (UCLA) has introduced a novel framework for designing and creating universal diffractive waveguides that can control the flow of light in highly specific and complex ways.

This new technology uses (AI), specifically deep learning, to design a series of structured surfaces that guide light with high efficiency and can perform a wide range of functions that are challenging for conventional waveguides.

The work is published in the journal Nature Communications.

Read more | >