“You learn from your mistakes”

๐Ÿ‡ฌ๐Ÿ‡ง You learn from your mistakes
๐Ÿ‡ป๐Ÿ‡ฆ Errando discitur
๐Ÿ‡ฎ๐Ÿ‡น Sbagliando si impara
๐Ÿ‡ฉ๐Ÿ‡ช Aus Fehlern lernt man
๐Ÿ‡ช๐Ÿ‡ธ Echando a perder se aprende
๐Ÿ‡ฏ๐Ÿ‡ต ไธƒ่ปขใณๅ…ซ่ตทใ โ€” Fall 7 times, stand up 8
๐Ÿ‡จ๐Ÿ‡ณ ๅคฑ่ดฅๆ˜ฏๆˆๅŠŸไน‹ๆฏ โ€” Failure is the mother of success

But guess who takes this more literally than anyone? Machine Learning algorithms.

MAKING MISTAKES…
The algorithm makes a prediction and calculates how wrong it was:

L = 1/n โˆ‘(y โˆ’ ลท)ยฒ

Translation: how far off were you from the right answer? Square it so it hurts even more. ๐Ÿ˜…

…AND LEARNING!
Parameters are adjusted in the direction that reduces the error:

ฮธ = ฮธ โˆ’ ฮท โˆ‚L/โˆ‚ฮธ

Translation: ok, next time I’ll correct a little. ฮท is “how much” I correct.

๐Ÿ” REPEAT a few million times.

From Neural Networks to Gradient Boosting, from Linear Regression to SVMs, the core idea is the same: measure your mistake, learn from it, try again.

A universal principle, from ancient proverbs to artificial intelligence.