Katie Hafner wrote this obituary of Arianna Rosenbluth, original programmer of what is known as the Metropolis algorithm:
Arianna Rosenbluth Dies at 93; Pioneering Figure in Data Science
Dr. Rosenbluth, who received her physics Ph.D. at 21, helped create an algorithm that has became a foundation of understanding huge quantities of data. She died of complications of the coronavirus. . . .
Despite her extraordinary work and despite earning her Ph.D. from Harvard at 21, Dr. Rosenbluth left the field in her mid-20s and rarely talked about her scientific achievement afterward. . . .
Arianna Wright was born on Sept. 15, 1927, in Houston to Augustus and Leffie (Woods) Wright. Her mother was a schoolteacher, and her father was an office manager for a flower company. . . .
Dr. Rosenbluth was an accomplished fencer, winning not only women’s championships but also men’s. But her plans to compete in the Olympics were foiled, first by the cancellation of the 1944 summer Games during World War II, then by a lack of funds for travel to the London Games in 1948, her daughter said. . . .
After completing her dissertation, she traveled west for a postdoctoral fellowship at Stanford University, funded by the Atomic Energy Commission. There she met Marshall Rosenbluth, another physicist. They married in 1951 and moved to New Mexico to work at Los Alamos. . . .
The group published its seminal paper, “Equation of State Calculations by Fast Computing Machines,” in The Journal of Chemical Physics in 1953. . . .
Arianna Rosenbluth’s contribution was crucial. “She actually did all the coding, which at that time was a new art for these new machines,” Marshall Rosenbluth said in 2003. Sophisticated programming tools were still years away, so Dr. Rosenbluth programmed in machine language . . .
Metropolis the man was involved in the Monte Carlo method in 1946, but the “Metropolis algorithm” (I guess we should call it the Rosenbluth and Rosenbluth algorithm) was from the early 1950s, at least that’s what Mr. Rosenbluth said. Metropolis in a 1987 memoir attributed the Monte Carlo algorithm to Ulam and Teller in 1947 or maybe 1946: they developed it for their H-bomb work. He also said that Fermi had independently discovered the principle in the early 1930s. Then again, maybe Laplace had some idea of this method back around 1800. The method becomes more useful once computing is available. This Metropolis article also has an amazing picture of a physical randomization device that was used to perform Monte Carlo simulations:
In that article, Metropolis mentions the method that is now called the Metropolis algorithm: “During this study a strategy was developed that led to greater computing efficiency for equilibrium systems obeying the Boltzmann distribution function. According to this strategy, if a statistical ‘move’ of a particle in the system resulted in a decrease in the energy of the system, the new configuration was accepted. On the other hand, if there was an increase in energy, the new configuration was accepted only if it survived a game of chance biased by a Boltzmann factor. Otherwise, the old configuration became a new statistic.” He describes this as “a collaborative effort with the Tellers, Edward and Mici, and the Rosenbluths, Marshall and Arianna.” But, as we know, Marshall said that Metropolis and Mici Teller didn’t do anything on it at all. So maybe this was just Metropolis being a diplomatic lab director.
In any case, the contributions of the Rosenbluths have endured. An algorithm becomes real when it is programmed.
And yet another entry in Stigler’s Law of Eponymy. Note that this example was already in the extensive Wikipedia list: https://en.wikipedia.org/wiki/List_of_examples_of_Stigler%27s_law#M
Thank you for sharing this. I’m always struck by the lack of depth of my understanding around the development and history of equations and methodologies I use every day. I’d love to read more about the history if anyone has any relatively succinct references.
David Salsburg’s relatively short book “The Lady Tasting Tea” has a good number of stories about the people and personalities involved in the development of modern statistics over the late 19th and 20th centuries. I’ve always been curious why we emphasize the stories of creators in teaching art, music, and even science but rarely seem to do so with math/statistics.
“This Metropolis article also has an amazing picture of a physical randomization device that was used to perform Monte Carlo simulations:” and accompanying diagram — Wow!!
Yeah! None for sale on ebay.
Fermi only ever built one, and it was found in somebodies abandoned office in 1966, and donated to the Bradbury Science Museum in Los Alamos. This guy builds a replica using a 3D printer and laser cutter:
https://hackaday.io/project/16655-fermihack
“Dr. Rosenbluth was an accomplished fencer, winning not only women’s championships but also men’s. But her plans to compete in the Olympics were foiled…”
I’m guessing she would’ve enjoyed the humor in that statement.
+1
(But I didn’t notice it until you pointed it out — thanks!)
Actually, it may be incorrect to state that the Tellers did nothing about The Paper. Quoting from Augusta “Mici” Teller Wikipedia page,
“she wrote an initial version of the MANIAC I code for the first paper introducing Markov chain Monte Carlo simulation, though the final code used in the publication was written in entirety by Arianna Rosenbluth.”
and still on Wikipedia,
“Rosenbluth credits Teller with a crucial but early suggestion to “take advantage of statistical mechanics and take ensemble averages instead of following detailed kinematics”.”
X:
Bunch of unreliable narrators, these physicists are!
No, just a many worlds interpretation.
I wonder what would happen if I start writing Arianna* algorithm (footnote: *also known as Metropolis algorithm) in my papers since today. What would probably be the editor’s response?
Mikhail:
The algorithm is Marshall’s, the code is Arianna’s. At least that’s what they say.
Don’t just speculate; try it and see!
Still Metropolis algorithm. “The original name will stick.”
From “Marshall Rosenbluth and the Metropolis algorithm” by J. E. Gubernatis:
Obviously, Marshall’s revelations prompted hallway and diner discussions about whether the Metropolis algorithm
should be called the Rosenbluth or at least the Rosenbluth–Teller algorithm. One of the organizing committee, Rajan
Gupta, whose curiosity got the best of him, privately asked Marshall his opinion about the name of the algorithm. Marshall replied:
“… life has been good to me. I feel rewarded in knowing that this algorithm will allow scientists to solve problems ranging from fluid flow to social dynamics to elucidating the nature of elementary particles.”
But why should we just believe what Rosenbluth said in 2003 when “Nick M. and Mici Teller are dead, Edward is unable to travel and has little remaining memory.”
“Marshall Rosenbluth and the Metropolis algorithm” by J. E. Gubernatis
https://www.jordanellenberg.com/book/shape/ says to the contrary that JSE is “himself a world class geometer”, which doesn’t happen without strong visualization ability (even in blind people).
Presumably that’s the publisher or publicist mistaking “arithmetic geometry” for a form of the visuospatial exercise known to the public as geometry. When in fact it’s a grownup version of Ellenberg’s nuke-it-with-algebra story, so that what he says he did in high school competitions, and what he does professionally are a similar sort of visualization avoidance.
sorry, wrong thread. That comment was for the Jordan Ellenberg SHAPE discussion in my other browser tab. I will repost there, moderator should feel free to delete here.
“Nuke it with algebra” . . . well put!