def test_autogenes_run_custom_objectives():
ag = AutoGeneS(np.random.randn(2, 5))
# Lambda as objective
obj1 = lambda x: len(x)
ag.run(ngen=0,
weights=(1, 1),
objectives=(obj1, 'distance'),
verbose=False)
assert ag.objectives_func == [obj1, ga_objectives.distance]
assert ag.objectives_names == ['<lambda>', 'distance']
# Invalid objective
def invalid_objective(data):
return data
with pytest.raises(Exception):
ag.run(ngen=0,
weights=(-1, 1),
objectives=(invalid_objective, 'distance'),
verbose=False)
# Normal function
def count(data):
return data.shape[0]
ag.run(ngen=0,
weights=(-1, 1),
objectives=(count, 'distance'),
verbose=False)
assert ag.objectives_func == [count, ga_objectives.distance]
assert ag.objectives_names == ['count', 'distance']
def ag_simple(): data = np.zeros((3,6)) data[0,0:2] = 1 data[1,2:4] = 1 data[2,4:6] = 1 return AutoGeneS(data)
def test_autogenes_init():
with pytest.raises(Exception):
AutoGeneS()
with pytest.raises(Exception):
AutoGeneS([[11, 1], [1]])
with pytest.raises(ValueError,
match="data is expected to have two dimensions"):
AutoGeneS(np.array([1]))
with pytest.raises(ValueError,
match="At least two rows \\(cell types\\) expected"):
AutoGeneS(np.int_([[1, 2, 3]]))
with pytest.raises(
ValueError,
match=
"Number of columns \\(genes\\) must be >= number of rows \\(cell types\\)"
):
AutoGeneS(np.int_(np.zeros((5, 2))))
with pytest.raises(
ValueError,
match=
"Number of columns \\(genes\\) must be >= number of rows \\(cell types\\)"
):
AutoGeneS(np.int_(np.zeros((2, 1))))
with pytest.raises(ValueError,
match="Some entries in data are not scalars"):
AutoGeneS(np.array([[1, np.inf], [1, np.nan]]))
arr1 = np.float64([[-1, 12.5], [1, 1E10]])
ag1 = AutoGeneS(arr1)
assert ag1.selection == None
assert np.array_equal(ag1.data, arr1)
def test_autogenes_run_objectives_and_weights():
ag = AutoGeneS(np.random.randn(2, 5))
ag.run()
assert ag.objectives_func == [
ga_objectives.correlation, ga_objectives.distance
]
with pytest.raises(Exception, match="Need objectives for weights"):
ag.run(weights=(1, 1))
with pytest.raises(Exception, match="Need weights for objectives"):
ag.run(objectives=('distance', 'test'))
with pytest.raises(ValueError, match="No such objective: test"):
ag.run(weights=(1, 1), objectives=('distance', 'test'))
with pytest.raises(ValueError, match="Invalid objective"):
ag.run(weights=(1, 1), objectives=(1, 1))
with pytest.raises(
ValueError,
match="Number of weights does not match number of objectives"):
ag.run(weights=(3, 4, 5), objectives=('distance', 'correlation'))
try:
ag.run(weights=(1, ), objectives=('distance', ), verbose=False)
except Exception:
pytest.fail("Should support single objective")
with pytest.raises(Exception):
ag.run(weights=1, objectives=('distance', ))
with pytest.raises(Exception):
ag.run(weights=('a', ), objectives=('distance', ))
with pytest.warns(UserWarning):
ag.run(weights=(3, 0, 5),
objectives=('distance', 'num_genes', 'correlation'))
assert ag.weights == (3, 5)
assert ag.objectives_func == [
ga_objectives.distance, ga_objectives.correlation
]
assert ag.objectives_names == ['distance', 'correlation']
def test_autogenes_run_general_args():
ag = AutoGeneS(np.random.randn(2, 5))
# mode = standard
ag.run(ngen=0, mode='standard', verbose=False)
# mode = fixed
with pytest.raises(ValueError, match="You need to supply nfeatures"):
ag.run(mode='fixed')
with pytest.raises(
ValueError,
match=
"nfeatures doesn't apply to standard mode \\(did you mean mode='fixed'\\?\\)"
):
ag.run(nfeatures=10)
with pytest.raises(
ValueError,
match="nfeatures must be <= the number of columns \\(genes\\)"):
ag.run(mode='fixed', nfeatures=10)
with pytest.raises(
ValueError,
match="nfeatures must be >= the number of rows \\(cell types\\)"):
ag.run(mode='fixed', nfeatures=1)
ag.run(ngen=0, mode='fixed', nfeatures=2, verbose=False)
with pytest.raises(ValueError, match="No such objective"):
ag.run(mode='fixed',
nfeatures=3,
weights=(1, 1),
objectives=('a', 'b'))
def ag_random(): np.random.seed(0) data = np.random.randint(0,5,(3,10)) return AutoGeneS(data)
def ag(): ag = AutoGeneS(np.identity(10)[:5]) return ag