Caution

You're reading an old version of this documentation. If you want up-to-date information, please have a look at 0.10.2.

librosa.util.tiny

librosa.util.tiny(x)[source]

Compute the tiny-value corresponding to an input’s data type.

This is the smallest “usable” number representable in x.dtype (e.g., float32).

This is primarily useful for determining a threshold for numerical underflow in division or multiplication operations.

Parameters:
xnumber or np.ndarray

The array to compute the tiny-value for. All that matters here is x.dtype

Returns:
tiny_valuefloat

The smallest positive usable number for the type of x. If x is integer-typed, then the tiny value for np.float32 is returned instead.

See also

numpy.finfo

Examples

For a standard double-precision floating point number:

>>> librosa.util.tiny(1.0)
2.2250738585072014e-308

Or explicitly as double-precision

>>> librosa.util.tiny(np.asarray(1e-5, dtype=np.float64))
2.2250738585072014e-308

Or complex numbers

>>> librosa.util.tiny(1j)
2.2250738585072014e-308

Single-precision floating point:

>>> librosa.util.tiny(np.asarray(1e-5, dtype=np.float32))
1.1754944e-38

Integer

>>> librosa.util.tiny(5)
1.1754944e-38