Toggle light / dark theme

A thermodynamic approach to machine learning: How optimal transport theory can improve generative models

Joint research led by Sosuke Ito of the University of Tokyo has shown that nonequilibrium thermodynamics, a branch of physics that deals with constantly changing systems, explains why optimal transport theory, a mathematical framework for the optimal change of distribution to reduce cost, makes generative models optimal. As nonequilibrium thermodynamics has yet to be fully leveraged in designing generative models, the discovery offers a novel thermodynamic approach to machine learning research. The findings were published in the journal Physical Review X.

Image generation has been improving in leaps and bounds over recent years: a video of a celebrity eating a bowl of spaghetti that represented the state of the art a couple of years ago would not even qualify as good today. The algorithms that power image generation are called diffusion models, and they contain randomness called “noise.”

During the training process, noise is introduced to the original data through diffusion dynamics. During the generation process, the model must eliminate the noise to generate new content from the noisy data. This is achieved by considering the time-reversed dynamics, as if playing the video in reverse. One piece of the art and science of building a model that produces high-quality content is specifying when and how much noise is added to the data.

Controlling polymer shapes: A new generation of shape-adaptive materials

What if a complex material could reshape itself in response to a simple chemical signal? A team of physicists from the University of Vienna and the University of Edinburgh has shown that even small changes in pH value and thus in electric charge can shift the spatial arrangement of closed ring-shaped polymers (molecular chains)—by altering the balance between twist and writhe, two distinct modes of spatial deformation.

Their findings, published in Physical Review Letters, demonstrate how electric charge can be used to reshape polymers in a reversible and controllable way—opening up new possibilities for programmable, responsive materials.

With such materials, permeability and such as elasticity, yield stress and viscosity could be better controlled and precisely “programmed.”

Stitched for strength: The physics of jamming in stiff, knitted fabrics

School of Physics Associate Professor Elisabetta Matsumoto is unearthing the secrets of the centuries-old practice of knitting through experiments, models, and simulations. Her goal? Leveraging knitting for breakthroughs in advanced manufacturing—including more sustainable textiles, wearable electronics, and soft robotics.

Matsumoto, who is also a principal investigator at the International Institute for Sustainability with Knotted Chiral Meta Matter (WPI-SKCM2) at Hiroshima University, is the corresponding author on a new study exploring the physics of ‘’—a phenomenon when soft or stretchy materials become rigid under low stress but soften under higher tension.

The study, “Pulling Apart the Mechanisms That Lead to Jammed Knitted Fabrics,” is published in Physical Review E, and also includes Georgia Tech Matsumoto Group graduate students Sarah Gonzalez and Alexander Cachine in addition to former postdoctoral fellow Michael Dimitriyev, who is now an assistant professor at Texas A&M University.

Astronomers Finally Crack the Mystery of Cosmic X-Ray Blasts

The mysterious cosmic explosion has been traced to a massive stellar explosion. A global team of astrophysicists, led by researchers from Northwestern University and the University of Leicester (England), may have identified the origin of fast X-ray transients (FXTs), mysterious, fleeting bursts of

Researchers develop flexible fiber material for self-powered health-monitoring sensors

Could clothing monitor a person’s health in real time, because the clothing itself would be a self-powered sensor? A new material created through electrospinning, which is a process that draws out fibers using electricity, brings this possibility one step closer.

A team led by researchers at Penn State has developed a new fabrication approach that optimizes the internal structure of electrospun fibers to improve their performance in electronic applications. The team has published its findings in the Journal of Applied Physics.

This novel electrospinning approach could open the door to more efficient, flexible and scalable electronics for wearable sensors, health monitoring and sustainable energy harvesting, according to Guanchun Rui, a visiting postdoctoral student in the Department of Electrical Engineering and the Materials Research Institute and co-lead author of the study.

Study finds cell cytoskeleton mimics critical phenomena seen in earthquakes and metals

Prof. Michael Murrell’s group (lead author Zachary Gao Sun, graduate student in physics) in collaboration with Prof. Garegin Papoian’s group from the University of Maryland at College Park has found critical phenomena (self-organized criticality) that are reminiscent of the earthquakes and avalanches inside the cell cytoskeleton through self-organization of purified protein components.

In a groundbreaking discovery, researchers have found that the cell’s cytoskeleton—the mechanical machinery of the cell—behaves much like Earth’s crust, constantly regulating how it dissipates energy and transmits information. This self-regulating behavior enables cells to carry out complex processes such as migration and division with remarkable precision.

Even more striking, the study draws parallels between the behavior of microscopic cellular structures and massive celestial bodies, suggesting that the principles of criticality—where systems naturally tune themselves to the brink of transformation—may be universal across vastly different scales of nature.

Physicists show tensor mesons play important role in light-on-light scattering

Usually, light waves can pass through each other without any resistance. According to the laws of electrodynamics, two light beams can exist in the same place without influencing each other; they simply overlap. Lightsaber battles, as seen in science fiction films, would therefore be rather boring in reality.

New framework clears spin-orbit confusion in solids and unifies physics

The researchers came up with a new way to describe how an electron’s spin interacts with the material it moves through, without using the complicated and unreliable tool called the orbital angular momentum operator, which usually causes problems in crystals.

Instead, they introduced a new idea called relativistic spin-lattice interaction. This basically means they focused on how an electron’s spin reacts to the structure of the solid itself, using principles from Einstein’s theory of relativity.

/* */