The Accumulator class is a data type for computing the running mean, sample standard deviation, and sample variance of a stream of real numbers. It provides an example of a mutable data type and a streaming algorithm.

This implementation uses a one-pass algorithm that is less susceptible to floating-point roundoff error than the more straightforward implementation based on saving the sum of the squares of the numbers. This technique is due to B. P. Welford. Each operation takes constant time in the worst case. The amount of memory is constant - the data values are not stored.