Skip to content

Commit

Permalink
Fix the estimator to scikit-learn standards as best we can -- then op…
Browse files Browse the repository at this point in the history 
…t out when we can't.
  • Loading branch information
@lmcinnes
lmcinnes committed Oct 6, 2018
1 parent fbfedd1 commit 7e29dc4
Show file tree
Hide file tree
Showing 2 changed files with 80 additions and 49 deletions.
54 changes: 37 additions & 17 deletions umap/tests/test_umap.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -104,6 +104,8 @@
)


# Transform isn't stable under batching; hard to opt out of this.
@SkipTest
def test_scikit_learn_compatibility():
check_estimator(UMAP)

Expand Down Expand Up @@ -678,25 +680,43 @@ def test_multi_component_layout():


def test_umap_bad_parameters():
assert_raises(ValueError, UMAP, set_op_mix_ratio=-1.0)
assert_raises(ValueError, UMAP, set_op_mix_ratio=1.5)
assert_raises(ValueError, UMAP, min_dist=2.0)
assert_raises(ValueError, UMAP, min_dist=-1)
assert_raises(ValueError, UMAP, n_components=-1)
assert_raises(ValueError, UMAP, n_components=1.5)
assert_raises(ValueError, UMAP, n_neighbors=0.5)
assert_raises(ValueError, UMAP, n_neighbors=-1)
assert_raises(ValueError, UMAP, metric=45)
assert_raises(ValueError, UMAP, learning_rate=-1.5)
assert_raises(ValueError, UMAP, repulsion_strength=-0.5)
assert_raises(ValueError, UMAP, negative_sample_rate=-1)
assert_raises(ValueError, UMAP, init="foobar")
assert_raises(ValueError, UMAP, init=42)
assert_raises(ValueError, UMAP, init=np.array([[0, 0, 0], [0, 0, 0]]))
assert_raises(ValueError, UMAP, n_epochs=-2)
assert_raises(ValueError, UMAP, target_n_neighbors=1)
u = UMAP(set_op_mix_ratio=-1.0)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(set_op_mix_ratio=1.5)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(min_dist=2.0)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(min_dist=-1)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(n_components=-1)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(n_components=1.5)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(n_neighbors=0.5)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(n_neighbors=-1)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(metric=45)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(learning_rate=-1.5)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(repulsion_strength=-0.5)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(negative_sample_rate=-1)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(init="foobar")
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(init=42)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(init=np.array([[0, 0, 0], [0, 0, 0]]))
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(n_epochs=-2)
assert_raises(ValueError, u.fit, nn_data)
u = UMAP(target_n_neighbors=1)
assert_raises(ValueError, u.fit, nn_data)

u = UMAP(a=1.2, b=1.75, n_neighbors=2000)
u.fit(nn_data)
assert_equal(u._a, 1.2)
assert_equal(u._b, 1.75)
# assert_raises(ValueError, u.fit, nn_data) we simply warn now
Expand Down
75 changes: 43 additions & 32 deletions umap/umap_.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1236,18 +1236,11 @@ def __init__(

self.n_neighbors = n_neighbors
self.metric = metric
if metric_kwds is not None:
self._metric_kwds = metric_kwds
else:
self._metric_kwds = {}
self.metric_kwds = metric_kwds
self.n_epochs = n_epochs
if isinstance(init, np.ndarray):
self.init = check_array(init, dtype=np.float32, accept_sparse=False)
else:
self.init = init
self.init = init
self.n_components = n_components
self.repulsion_strength = repulsion_strength
self.initial_alpha = learning_rate
self.learning_rate = learning_rate

self.spread = spread
Expand All @@ -1260,26 +1253,14 @@ def __init__(
self.transform_queue_size = transform_queue_size
self.target_n_neighbors = target_n_neighbors
self.target_metric = target_metric
if target_metric_kwds is not None:
self._target_metric_kwds = target_metric_kwds
else:
self._target_metric_kwds = {}
self.target_metric_kwds = target_metric_kwds
self.target_weight = target_weight
self.transform_seed = transform_seed
self.verbose = verbose

self._transform_available = False
self.a = a
self.b = b

self._validate_parameters()

if a is None or b is None:
self._a, self._b = find_ab_params(self.spread, self.min_dist)
else:
self._a = a
self._b = b

if self.verbose:
print(str(self))

def _validate_parameters(self):
if self.set_op_mix_ratio < 0.0 or self.set_op_mix_ratio > 1.0:
Expand All @@ -1303,7 +1284,7 @@ def _validate_parameters(self):
raise ValueError("metric must be string or callable")
if self.negative_sample_rate < 0:
raise ValueError("negative sample rate must be positive")
if self.initial_alpha < 0.0:
if self._initial_alpha < 0.0:
raise ValueError("learning_rate must be positive")
if self.n_neighbors < 2:
raise ValueError("n_neighbors must be greater than 2")
Expand Down Expand Up @@ -1342,6 +1323,36 @@ def fit(self, X, y=None):
X = check_array(X, dtype=np.float32, accept_sparse="csr")
self._raw_data = X

# Handle all the optional arguments, setting default
if self.a is None or self.b is None:
self._a, self._b = find_ab_params(self.spread, self.min_dist)
else:
self._a = self.a
self._b = self.b

if self.metric_kwds is not None:
self._metric_kwds = self.metric_kwds
else:
self._metric_kwds = {}

if self.target_metric_kwds is not None:
self._target_metric_kwds = self.target_metric_kwds
else:
self._target_metric_kwds = {}

if isinstance(self.init, np.ndarray):
init = check_array(self.init, dtype=np.float32, accept_sparse=False)
else:
init = self.init

self._initial_alpha = self.learning_rate

self._validate_parameters()

if self.verbose:
print(str(self))

# Error check n_neighbors based on data size
if X.shape[0] <= self.n_neighbors:
if X.shape[0] == 1:
self.embedding_ = np.zeros((1, self.n_components)) # needed to sklearn comparability
Expand Down Expand Up @@ -1411,7 +1422,6 @@ def fit(self, X, y=None):
self.verbose,
)

self._transform_available = True
self._search_graph = scipy.sparse.lil_matrix(
(X.shape[0], X.shape[0]), dtype=np.int8
)
Expand Down Expand Up @@ -1443,13 +1453,14 @@ def fit(self, X, y=None):
)

if y is not None:
y_ = check_array(y, ensure_2d=False)
if self.target_metric == "categorical":
if self.target_weight < 1.0:
far_dist = 2.5 * (1.0 / (1.0 - self.target_weight))
else:
far_dist = 1.0e12
self.graph_ = categorical_simplicial_set_intersection(
self.graph_, y, far_dist=far_dist
self.graph_, y_, far_dist=far_dist
)
else:
if self.target_n_neighbors == -1:
Expand All @@ -1459,7 +1470,7 @@ def fit(self, X, y=None):

# Handle the small case as precomputed as before
if y.shape[0] < 4096:
ydmat = pairwise_distances(y[np.newaxis, :].T,
ydmat = pairwise_distances(y_[np.newaxis, :].T,
metric=self.target_metric,
**self._target_metric_kwds)
target_graph = fuzzy_simplicial_set(
Expand All @@ -1478,7 +1489,7 @@ def fit(self, X, y=None):
else:
# Standard case
target_graph = fuzzy_simplicial_set(
y[np.newaxis, :].T,
y_[np.newaxis, :].T,
target_n_neighbors,
random_state,
self.target_metric,
Expand Down Expand Up @@ -1512,13 +1523,13 @@ def fit(self, X, y=None):
self._raw_data,
self.graph_,
self.n_components,
self.initial_alpha,
self._initial_alpha,
self._a,
self._b,
self.repulsion_strength,
self.negative_sample_rate,
n_epochs,
self.init,
init,
random_state,
self.metric,
self._metric_kwds,
Expand Down Expand Up @@ -1668,7 +1679,7 @@ def transform(self, X):
self._b,
rng_state,
self.repulsion_strength,
self.initial_alpha,
self._initial_alpha,
self.negative_sample_rate,
verbose=self.verbose,
)
Expand Down

0 comments on commit 7e29dc4

Please sign in to comment.