This talk took us on a tour through one of the most quietly powerful equations in mathematics: the heat equation. On the surface, it’s “just” a formula describing how temperature spreads through space over time. In reality, it turns out to be a kind of mathematical celebrity, popping up everywhere from physics to finance.
The talk began by unpacking what the equation says. Temperature at a point changes depending on how curved the temperature is around it – in other words, heat flows from hotter regions to colder ones, smoothing things out. Partial derivatives made a cameo appearance here, framed intuitively as rates of change and curvature rather than scary symbols.
Then came some history. Long before the equation became standard, heat was thought of as a weightless fluid (“caloric”). That changed thanks to Joseph Fourier, whose radical claim that any function – even jagged ones – could be written as a sum of sines and cosines initially got his work rejected. Awkward, given how foundational it later became.
One of the most satisfying moments was seeing why the Gaussian curve appears when you solve the heat equation. Start with a single point of heat, let time run, and out pops the familiar bell curve – not by coincidence, but because energy spreads out evenly while staying conserved. This same mathematics underpins diffusion, Brownian motion, and even the Black–Scholes equation in finance.
The talk ended by confronting reality: the heat equation is notoriously hard to simulate. Computers don’t “do” calculus, and higher-dimensional shapes make things worse. The solution? Physics-informed neural networks, which cleverly bake the equation itself into a machine-learning model.
From desert heat to neural networks, this was a great
reminder that elegant equations often have the widest reach.
- Louis