normal distribution

The normal distribution is the famous bell curve: a symmetric hump that piles most of its probability near the middle and tapers off smoothly toward both extremes. Heights, measurement errors, exam scores, and countless other natural quantities scatter into roughly this shape — clustered around a typical value, with the unusually large and unusually small both growing rarer the farther out you go.

Just two numbers pin it down completely. The mean fixes where the peak sits — the center of mass. The standard deviation sets how wide and flat the bell is — a small one gives a tall narrow spike (everything tightly bunched), a large one gives a low broad mound (values spread far and wide). A handy rule of thumb: about 68% of the data falls within one standard deviation of the mean, about 95% within two, and about 99.7% within three.

It dominates statistics and machine learning for a deep reason called the central limit theorem: when you add up lots of small, independent random effects, the total tends toward a normal shape almost regardless of where the pieces came from. So errors, noise, and aggregates naturally land here. Models lean on it constantly — as the assumed shape of noise, as the starting random spread for a network's weights, as the bell in a Gaussian mixture. The honest warning: plenty of real data is not normal at all. Incomes, city sizes, and stock crashes have fat tails where rare extremes happen far more often than a bell curve predicts, and assuming normality there can be dangerously misleading.