“You learn from your mistakes”
๐ฌ๐ง You learn from your mistakes
๐ป๐ฆ Errando discitur
๐ฎ๐น Sbagliando si impara
๐ฉ๐ช Aus Fehlern lernt man
๐ช๐ธ Echando a perder se aprende
๐ฏ๐ต ไธ่ปขใณๅ
ซ่ตทใ โ Fall 7 times, stand up 8
๐จ๐ณ ๅคฑ่ดฅๆฏๆๅไนๆฏ 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 โ(y โ ลท)ยฒ
Translation: how far off were you from the right answer? Square it so it hurts even more. ๐
…AND LEARNING!
Parameters are adjusted in the direction that reduces the error:
ฮธ = ฮธ โ ฮท โL/โฮธ
Translation: ok, next time I’ll correct a little. ฮท is “how much” I correct.
๐ 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.








