Santiago Piccinini spiccinini

## anonymous
"""Estimate parameters of a Gaussian from a histogram.

You typically get about 10% error with a sample size of 32 in 8 bins
that cover its range reasonably well.  The errors look pretty Gaussian
with some weird rounding artifacts.

For 8 bins, this algorithm requires 16 multiplications, 8 additional
squarings, 28 additions, a subtraction, a threshold to zero, two divisions,
and a square root, about 50 numerical operations in all.  But the operations
are easily vectorizable.  Also they can be all executed in a single pass over
	"""Estimate parameters of a Gaussian from a histogram.

	You typically get about 10% error with a sample size of 32 in 8 bins
	that cover its range reasonably well. The errors look pretty Gaussian
	with some weird rounding artifacts.

	For 8 bins, this algorithm requires 16 multiplications, 8 additional
	squarings, 28 additions, a subtraction, a threshold to zero, two divisions,
	and a square root, about 50 numerical operations in all. But the operations
	are easily vectorizable. Also they can be all executed in a single pass over