def test_3_1():
s = SearchSpace()
s.add_float('b', -2, 2)
values = [s.rvs()['b'] for _ in range(100)]
assert all(-2 < v < 2 for v in values)
_run_chi2_test(values, bin_edges=np.linspace(-2, 2, 10))
def test_3_1():
s = SearchSpace()
s.add_float('b', -2, 2)
values = [s.rvs()['b'] for _ in range(100)]
assert all(-2 < v < 2 for v in values)
_run_chi2_test(values, bin_edges=np.linspace(-2, 2, 10))
def test_moe_rest_1():
moe_url = os.environ.get('MOE_API_URL', 'http://ERROR-sdjssfssdbsdf.com')
try:
request.urlopen(moe_url)
except error.URLError:
raise nose.SkipTest(
'No available MOE REST API endpoint (set with '
'MOE_API_URL environment variable)')
searchspace = SearchSpace()
searchspace.add_float('x', -10, 10)
searchspace.add_float('y', 1, 10, warp='log')
searchspace.add_int('z', -10, 10)
searchspace.add_enum('w', ['opt1', 'opt2'])
history = [(searchspace.rvs(), np.random.random(), 'SUCCEEDED')
for _ in range(4)]
params = MOE(url=moe_url).suggest(history, searchspace)
for k, v in iteritems(params):
assert k in searchspace.variables
if isinstance(searchspace[k], EnumVariable):
assert v in searchspace[k].choices
elif isinstance(searchspace[k], FloatVariable):
assert searchspace[k].min <= v <= searchspace[k].max
elif isinstance(searchspace[k], IntVariable):
assert searchspace[k].min <= v <= searchspace[k].max
else:
assert False
def test_3_2():
s = SearchSpace()
s.add_float('b', -2, 2)
values = [pyll.stochastic.sample(s['b'].to_hyperopt())
for _ in xrange(100)]
assert all(-2 < v < 2 for v in values)
_run_chi2_test(values, bin_edges=np.linspace(-2, 2, 10))
def test_3_2():
s = SearchSpace()
s.add_float('b', -2, 2)
values = [pyll.stochastic.sample(s['b'].to_hyperopt())
for _ in xrange(100)]
assert all(-2 < v < 2 for v in values)
_run_chi2_test(values, bin_edges=np.linspace(-2, 2, 10))
def test_5_2():
s = SearchSpace()
s.add_float('a', 1e-5, 1, warp='log')
bin_edges = np.logspace(np.log10(s['a'].min), np.log10(s['a'].max), num=5)
values = [pyll.stochastic.sample(s['a'].to_hyperopt())
for _ in xrange(100)]
_run_chi2_test(values, bin_edges)
def test_5_2():
s = SearchSpace()
s.add_float('a', 1e-5, 1, warp='log')
bin_edges = np.logspace(np.log10(s['a'].min), np.log10(s['a'].max), num=5)
values = [pyll.stochastic.sample(s['a'].to_hyperopt())
for _ in xrange(100)]
_run_chi2_test(values, bin_edges)
def test_5_1():
s = SearchSpace()
s.add_float('a', 1e-5, 1, warp='log')
n_bins = 10
n_samples = 1000
bin_edges = np.logspace(np.log10(s['a'].min), np.log10(s['a'].max),
num=n_bins+1)
values = [s.rvs()['a'] for _ in xrange(n_samples)]
_run_chi2_test(values, bin_edges)
def test_5_1():
s = SearchSpace()
s.add_float('a', 1e-5, 1, warp='log')
n_bins = 10
n_samples = 1000
bin_edges = np.logspace(np.log10(s['a'].min), np.log10(s['a'].max),
num=n_bins+1)
values = [s.rvs()['a'] for _ in xrange(n_samples)]
_run_chi2_test(values, bin_edges)
def our_x2_iterates(n_iters=100):
history = []
searchspace = SearchSpace()
searchspace.add_float('x', -10, 10)
random = np.random.RandomState(0)
# note the switch of sign, because _our_ function hyperopt_tpe is
# a maximizer, not a minimizer
fn = lambda params: -params['x']**2
for i in range(n_iters):
params = HyperoptTPE(seed=random).suggest(history, searchspace)
history.append((params, fn(params), 'SUCCEEDED'))
return np.array([h[0]['x'] for h in history])
def our_x2_iterates(n_iters=100):
history = []
searchspace = SearchSpace()
searchspace.add_float('x', -10, 10)
random = np.random.RandomState(0)
# note the switch of sign, because _our_ function hyperopt_tpe is
# a maximizer, not a minimizer
def fn(params):
return -params['x']**2
for i in range(n_iters):
params = HyperoptTPE(seed=random).suggest(history, searchspace)
history.append((params, fn(params), 'SUCCEEDED'))
return np.array([h[0]['x'] for h in history])
def test_1():
s = SearchSpace()
s.add_int('a', 1, 2)
s.add_float('b', 2, 3)
s.add_enum('c', ['a', 'b', 'c'])
assert s.n_dims == 3
assert s['a'].min == 1
assert s['a'].max == 2
assert s['a'].name == 'a'
assert s['b'].min == 2
assert s['b'].max == 3
assert s['b'].name == 'b'
assert s['c'].choices == ['a', 'b', 'c']
assert s['c'].name == 'c'
def test_1():
s = SearchSpace()
s.add_int('a', 1, 2)
s.add_float('b', 2, 3)
s.add_enum('c', ['a', 'b', 'c'])
assert s.n_dims == 3
assert s['a'].min == 1
assert s['a'].max == 2
assert s['a'].name == 'a'
assert s['b'].min == 2
assert s['b'].max == 3
assert s['b'].name == 'b'
assert s['c'].choices == ['a', 'b', 'c']
assert s['c'].name == 'c'
def test_gp():
searchspace = SearchSpace()
searchspace.add_float('x', -10, 10)
searchspace.add_float('y', 1, 10, warp='log')
searchspace.add_int('z', -10, 10)
searchspace.add_enum('w', ['opt1', 'opt2'])
history = [(searchspace.rvs(), np.random.random(), 'SUCCEEDED')
for _ in range(4)]
params = GP().suggest(history, searchspace)
for k, v in iteritems(params):
assert k in searchspace.variables
if isinstance(searchspace[k], EnumVariable):
assert v in searchspace[k].choices
elif isinstance(searchspace[k], FloatVariable):
assert searchspace[k].min <= v <= searchspace[k].max
elif isinstance(searchspace[k], IntVariable):
assert searchspace[k].min <= v <= searchspace[k].max
else:
assert False
def test_gp():
searchspace = SearchSpace()
searchspace.add_float('x', -10, 10)
searchspace.add_float('y', 1, 10, warp='log')
searchspace.add_int('z', -10, 10)
searchspace.add_enum('w', ['opt1', 'opt2'])
history = [(searchspace.rvs(), np.random.random(), 'SUCCEEDED')
for _ in range(4)]
params = GP().suggest(history, searchspace)
for k, v in iteritems(params):
assert k in searchspace.variables
if isinstance(searchspace[k], EnumVariable):
assert v in searchspace[k].choices
elif isinstance(searchspace[k], FloatVariable):
assert searchspace[k].min <= v <= searchspace[k].max
elif isinstance(searchspace[k], IntVariable):
assert searchspace[k].min <= v <= searchspace[k].max
else:
assert False
def test_random():
searchspace = SearchSpace()
searchspace.add_float('x', -10, 10)
random = np.random.RandomState(0)
RandomSearch(seed=random).suggest([], searchspace)
def test_random():
searchspace = SearchSpace()
searchspace.add_float('x', -10, 10)
random = np.random.RandomState(0)
RandomSearch(seed=random).suggest([], searchspace)
