Andrey Tikhonov

Andrey Nikolayevich Tikhonov (1906–1993) was a Soviet mathematician who made his first major discovery in topology while still a teenager, then spent the rest of his career taming problems that seem, at first, to be unsolvable. His speciality was "ill-posed" problems — the frustrating ones where a tiny wobble in the data produces a wild, useless answer.

The idea that outlived them

Tikhonov's cure for these unstable problems was beautifully simple: don't just fit the data, also gently penalise crazy, oversized solutions. That trade-off is Tikhonov regularization, and it's the same idea machine-learning people call "ridge regression":

\min_{x}\; \|Ax - b\|^2 + \lambda\,\|x\|^2.

That extra \lambda\,\|x\|^2 term keeps the answer sensible when the raw problem would blow up. It's now a default tool in imaging, statistics and data science. His name also lives on in the Tychonoff theorem, a cornerstone of topology (his surname just gets spelled a few different ways in English).

If you go looking for him you'll find "Tikhonov," "Tychonoff," "Tichonov" and more — all the same man, just transliterated from the Russian by different hands over the decades. It leads to a fun quirk: students sometimes learn "Tychonoff's theorem" in a topology class and "Tikhonov regularization" in a data-science class without ever realising they were taught by the same person. He was also a serious academic organiser, helping to found a whole faculty of computational mathematics in Moscow.