3 ms·
> Machine learning, at its core, is about generalization, not memorization. Well they memorize the patterns. memorization doesnt mean rote learning.
by dominotw 20d ago
> Machine learning, at its core, is about generalization, not memorization.
Well they memorize the patterns.
memorization doesnt mean rote learning.
- porridgeraisin 20d agoThat's a bit pedantic no. Memorization in ML refers to the model having the wrong level of capacity such that it's too hard to optimise it such that it doesn't memorize the _training examples_ themselves.
- tomrod 20d agoI don't take issue with that. Attempting memorized pattern generalization through holdout / validation strategies is a big part of ML that you would not typically see with econometrics / psychometrics / possibly sabermetrics / most other -metrics. Philosophically the explain versus predict divide. https://www.stat.berkeley.edu/~aldous/157/Papers/shmueli.pdf https://www.stat.berkeley.edu/~aldous/157/Papers/shmueli.pdf
- red75prime 20d agoDoes '110001111000000011111111111' contain n>1∧∀d(d|n→(d=1∨d=n)) as well as infinite number of other generalizations?