def fit_transform(self, X, init=None, y=None):
"""
Fit the data from X, and returns the embedded coordinates
Parameters
----------
X : array, shape=[n_samples, n_features], or [n_samples, n_samples] \
if dissimilarity='precomputed'
Input data.
init : {None or ndarray, shape (n_samples,)}, optional
If None, randomly chooses the initial configuration
if ndarray, initialize the SMACOF algorithm with this array.
"""
if X.shape[0] == X.shape[1] and self.dissimilarity != "precomputed":
warnings.warn("The MDS API has changed. ``fit`` now constructs an"
" dissimilarity matrix from data. To use a custom "
"dissimilarity matrix, set "
"``dissimilarity=precomputed``.")
if self.dissimilarity == "precomputed":
self.dissimilarity_matrix_ = X
elif self.dissimilarity == "euclidean":
self.dissimilarity_matrix_ = euclidean_distances(X)
else:
raise ValueError("Proximity must be 'precomputed' or 'euclidean'."
" Got %s instead" % str(self.dissimilarity))
self.embedding_, self.stress_ = smacof(
self.dissimilarity_matrix_, metric=self.metric,
n_components=self.n_components, init=init, n_init=self.n_init,
n_jobs=self.n_jobs, max_iter=self.max_iter, verbose=self.verbose,
eps=self.eps, random_state=self.random_state)
return self.embedding_
def predict(self, features, ignore_first=False):
"""Predict from features, a numpy array of size (n_samples, n_features) Use the
training data to predict labels on the test features. For each testing sample, compare it
to the training samples. Look at the self.n_neighbors closest samples to the
test sample by comparing their feature vectors. The label for the test sample
is the determined by aggregating the K nearest neighbors in the training data.
Note that when using KNN for imputation, the predicted labels are the imputed testing data
and the shape is (n_samples, n_features).
Arguments:
features {np.ndarray} -- Features of each data point, shape of (n_samples,
n_features).
ignore_first {bool} -- If this is True, then we ignore the closest point
when doing the aggregation. This is used for collaborative
filtering, where the closest point is itself and thus is not a neighbor.
In this case, we would use 1:(n_neighbors + 1).
Returns:
labels {np.ndarray} -- Labels for each data point, of shape (n_samples,
n_dimensions). This n_dimensions should be the same as n_dimensions of targets in fit function.
"""
labels = np.zeros([len(features), self.n_dimensions])
for i in range(
0, len(features)
): # loops through each sample to find its closest neighbor
if self.distance_measure == 'euclidean':
distances = euclidean_distances(
self.features, features[i]
) # finds the distances between the testing set and the new sample
elif self.distance_measure == 'manhattan':
distances = manhattan_distances(self.features, features[i])
else:
distances = cosine_distances(self.features, features[i])
"""
print('Rows of self.features = %d' % len(self.features))
print('Columns of self.features = %d' % len(self.features[0]))
print('Columns of features[%d] = %d' % (i, len(features[i])))
"""
new_list = np.append(
distances, self.targets, axis=1
) # makes 1 big matrix with the 1st column being the distances and the other columns being their corresponding targets
new_list = new_list[np.argsort(
new_list[:, 0]
)] # sorts the matrix by the 1st column by ascending order (first rows have lowest value)
if ignore_first:
close_neighbors = new_list[1:self.n_neighbors +
1] # KNN is closest K neighbors
else:
close_neighbors = new_list[:self.n_neighbors]
char = np.empty(self.n_dimensions)
"""
if self.aggregator == 'mean':
char = statistics.mean(close_neighbors[:, 1:]) # finds mean of first column of closest neighbors
elif self.aggregator == 'mode':
char = statistics.mode(close_neighbors[:, 1:]) # finds mode of first column of closest neighbors
else:
char = statistics.median(close_neighbors[:, 1:]) # finds median of first column of closest neighbors
"""
for j in range(0, self.n_dimensions):
if self.aggregator == 'mean':
closest_neighbor = statistics.mean(
close_neighbors[:, j + 1]
) # finds mean of first column of closest neighbors
elif self.aggregator == 'mode':
closest_neighbor = statistics.mode(
close_neighbors[:, j + 1]
) # finds mode of first column of closest neighbors
else:
closest_neighbor = statistics.median(
close_neighbors[:, j + 1]
) # finds median of first column of closest neighbors
char[j] = closest_neighbor
labels[i, :] = char
#labels = np.append(labels, char, axis = 1)
return labels
def _smacof_single(similarities, metric=True, n_components=2, init=None,
max_iter=300, verbose=0, eps=1e-3, random_state=None):
"""
Computes multidimensional scaling using SMACOF algorithm
Parameters
----------
similarities: symmetric ndarray, shape [n * n]
similarities between the points
metric: boolean, optional, default: True
compute metric or nonmetric SMACOF algorithm
n_components: int, optional, default: 2
number of dimension in which to immerse the similarities
overwritten if initial array is provided.
init: {None or ndarray}, optional
if None, randomly chooses the initial configuration
if ndarray, initialize the SMACOF algorithm with this array
max_iter: int, optional, default: 300
Maximum number of iterations of the SMACOF algorithm for a single run
verbose: int, optional, default: 0
level of verbosity
eps: float, optional, default: 1e-6
relative tolerance w.r.t stress to declare converge
random_state: integer or numpy.RandomState, optional
The generator used to initialize the centers. If an integer is
given, it fixes the seed. Defaults to the global numpy random
number generator.
Returns
-------
X: ndarray (n_samples, n_components), float
coordinates of the n_samples points in a n_components-space
stress_: float
The final value of the stress (sum of squared distance of the
disparities and the distances for all constrained points)
"""
n_samples = similarities.shape[0]
random_state = check_random_state(random_state)
if similarities.shape[0] != similarities.shape[1]:
raise ValueError("similarities must be a square array (shape=%d)" %
n_samples)
if not np.allclose(similarities, similarities.T):
raise ValueError("similarities must be symmetric")
sim_flat = ((1 - np.tri(n_samples)) * similarities).ravel()
sim_flat_w = sim_flat[sim_flat != 0]
if init is None:
# Randomly choose initial configuration
X = random_state.rand(n_samples * n_components)
X = X.reshape((n_samples, n_components))
else:
# overrides the parameter p
n_components = init.shape[1]
if n_samples != init.shape[0]:
raise ValueError("init matrix should be of shape (%d, %d)" %
(n_samples, n_components))
X = init
old_stress = None
ir = IsotonicRegression()
for it in range(max_iter):
# Compute distance and monotonic regression
dis = euclidean_distances(X)
if metric:
disparities = similarities
else:
dis_flat = dis.ravel()
# similarities with 0 are considered as missing values
dis_flat_w = dis_flat[sim_flat != 0]
# Compute the disparities using a monotonic regression
disparities_flat = ir.fit_transform(sim_flat_w, dis_flat_w)
disparities = dis_flat.copy()
disparities[sim_flat != 0] = disparities_flat
disparities = disparities.reshape((n_samples, n_samples))
disparities *= np.sqrt((n_samples * (n_samples - 1) / 2) /
(disparities ** 2).sum())
# Compute stress
stress = ((dis.ravel() - disparities.ravel()) ** 2).sum() / 2
# Update X using the Guttman transform
dis[dis == 0] = 1e-5
ratio = disparities / dis
B = - ratio
B[np.arange(len(B)), np.arange(len(B))] += ratio.sum(axis=1)
X = 1. / n_samples * np.dot(B, X)
dis = np.sqrt((X ** 2).sum(axis=1)).sum()
if verbose >= 2:
print('it: %d, stress %s' % (it, stress))
if old_stress is not None:
if(old_stress - stress / dis) < eps:
if verbose:
print('breaking at iteration %d with stress %s' % (it,
stress))
break
old_stress = stress / dis
return X, stress
