Methods drawn from physics
The win is already turning heads on social media as a result of it appears uncommon that analysis in a pc science subject like machine studying would possibly win a Nobel Prize for physics. “And the 2024 Nobel Prize in Physics doesn’t go to physics…” tweeted German physicist Sabine Hossenfelder this morning.
From the Nobel committee’s point of view, the award largely derives from the truth that the 2 males drew from statistical fashions utilized in physics and partly from recognizing the developments in physics analysis that got here from utilizing the boys’s neural community strategies as analysis instruments.
Nobel committee chair Ellen Moons, a physicist at Karlstad College, Sweden, stated in the course of the announcement, “Synthetic neural networks have been used to advance analysis throughout physics subjects as various as particle physics, materials science and astrophysics.”
Hopfield, a 91-year-old theoretical biologist with a physics background, made a breakthrough in 1982 by creating a community that described connections between nodes as bodily forces, as Nature describes in a report. His innovation, referred to as the Hopfield network, makes use of ideas from physics that describe how atomic spins behave in supplies. Particularly, it shops patterns as low-energy states, permitting the system to recreate photos when prompted with related patterns. This strategy mimicked associative reminiscence, resembling how the mind recollects phrases or ideas.
Hinton, who’s 76, built upon Hopfield’s analysis within the early Nineteen Eighties by creating a layered model of the Hopfield community that integrated chances. Hinton drew parallels to physics research of huge methods of comparable parts like gasoline molecules. As a substitute of monitoring particular person molecules, physicists study collective properties like stress or temperature. The Boltzmann equation from Nineteenth-century physics calculates the likelihood of various states in such methods. Hinton utilized this idea to neural networks, naming his 1985 methodology the “Boltzmann machine,” which highlighted the connection between machine studying and statistical physics. A Boltzmann machine is able to recognizing and classifying photos and producing new examples primarily based on its coaching information.
