optunity.solvers.RandomSearch module

class optunity.solvers.RandomSearch.RandomSearch(num_evals, **kwargs)

Bases: optunity.solvers.util.Solver

Please refer to Random Search for more details about this algorithm.

Initializes the solver with bounds and a number of allowed evaluations. kwargs must be a dictionary of parameter-bound pairs representing the box constraints. Bounds are a 2-element list: [lower_bound, upper_bound].

>>> s = RandomSearch(x=[0, 1], y=[-1, 2], num_evals=50)
>>> s.bounds['x']
[0, 1]
>>> s.bounds['y']
[-1, 2]
>>> s.num_evals
50

bounds

Returns a dictionary containing the box constraints.

desc_brief = 'random parameter tuples sampled uniformly within box constraints'

desc_full = ['Tests random parameter tuples sampled uniformly within the box constraints.', ' ', 'This function requires the following arguments:', '- num_evals :: number of tuples to test', '- box constraints via keywords: constraints are lists [lb, ub]', ' ', 'This solver performs num_evals function evaluations.', ' ', 'This solver implements the technique described here:', 'Bergstra, James, and Yoshua Bengio. Random search for hyper-parameter optimization. Journal of Machine Learning Research 13 (2012): 281-305.']

lower

Returns the lower bound of par.

maximize(f, pmap=<built-in function map>)

Maximizes f.

Parameters:

• f (callable) – the objective function
• pmap (callable) – the map() function to use

Returns:

• the arguments which optimize f
• an optional solver report, can be None

minimize(f, pmap=<built-in function map>)

Minimizes f.

Parameters:

• f (callable) – the objective function
• pmap (callable) – the map() function to use

Returns:

• the arguments which optimize f
• an optional solver report, can be None

name = 'random search'

num_evals

Returns the number of evaluations this solver may do.

optimize(f, maximize=True, pmap=<built-in function map>)

Optimizes f.

Parameters:

• f (callable) – the objective function
• maximize (boolean) – do we want to maximizes?
• pmap (callable) – the map() function to use

Returns:

• the arguments which optimize f
• an optional solver report, can be None

static suggest_from_box(num_evals, **kwargs)

Creates a RandomSearch solver that uses num_evals evaluations within given bounds (lb, ub). The bounds are first tightened, resulting in new bounds covering 99% of the area.

>>> s = RandomSearch.suggest_from_box(30, x=[0, 1], y=[-1, 0], z=[-1, 1])
>>> s['x'] 
[0.005, 0.995]
>>> s['y'] 
[-0.995, -0.005]
>>> s['z'] 
[-0.99, 0.99]
>>> s['num_evals']
30

Verify that we can effectively make a solver from box.

>>> s = RandomSearch.suggest_from_box(30, x=[0, 1], y=[-1, 0], z=[-1, 1])
>>> solver = RandomSearch(**s)

upper

Returns the upper bound of par.