coopihc.base.Space.CatSet

class CatSet(array=None, seed=None, dtype=None, contains='numpy')[source]

Bases: coopihc.base.Space.BaseSpace

Categorical Set

A categorical set defined explicitly. Use this for data where traditional distance is meaningless i.e. when 1 is not closer to 0 then to 5. Performance of this object for large dimensions may be bad, because the whole array is stored in memory.

s = space(array=numpy.array([1, 2, 3], dtype=numpy.int16))
assert s.dtype == numpy.int16
assert s.N == 3
assert s.shape == ()

You can further set the seed of the space (useful when sampling from the space), force the dtype of the space and specify how membership to the space is checked via the keyword arguments. See BaseSpace for more information.

Parameters: array (numpy.ndarray, optional) – set of values, defaults to None

Methods

`sample`	Generate values by sampling uniformly from the set.
`serialize`

Attributes

`N`	Cardinality of the set
`dtype`	numpy.dtype of data
`high`
`low`
`shape`	numpy shape of the data that belongs to the set
`spacetype`

property N: Cardinality of the set

property dtype: numpy.dtype of data

sample()[source]

Generate values by sampling uniformly from the set. You can set the seed to the rng, see keyword arguments at init.

s = CatSet(array=numpy.arange(1000), seed=123)
q = CatSet(array=numpy.arange(1000), seed=123)
r = CatSet(array=numpy.arange(1000), seed=12)
_s, _q, _r = s.sample(), q.sample(), r.sample()
assert _s in s
assert _q in q
assert _r in r
assert _s == _q
assert _s != _r

property shape: numpy shape of the data that belongs to the set