This section provides an overview of the Triglav and the different parameters
of the Triglav class and its methods.
class triglav.Triglav(transformer = NoScale(), sampler = NoResample(), estimator = ExtraTreesClassifier(512, bootstrap = True),
stage_2_estimator = ExtraTreesClassifier(512, bootstrap = True), per_class_imp = False,
n_iter = 40, n_iter_fwer = 11, p_1 = 0.65, p_2 = 0.30, metric = "correlation", linkage = "complete",
thresh = 2.0, criterion = "distance", run_stage_2 = True, verbose = 0, n_jobs = 10)
transformer: default = NoScale()
The transformer to be used to scale features. One can use
the scikit-learn.preprocessing transformers. In addition,
CLR and Scaler (converts each row into frequencies) are
available by importing 'CLRTransformer' and 'Scaler' from the
'triglav' package.
sampler: default = NoResample()
The resampling method used for imbalanced classes. Should be
compatable with 'imblearn' or use an 'imblearn' resampler.
estimator: default = ExtraTreesClassifier(512, bootstrap = True)
The estimator used to calculate Shapley scores.
stage_2_estimator: default = ExtraTreesClassifier(512)
The estimator used to calculate SAGE values. Only used if the
'run_stage_2' is set to True.
per_class_imp: bool, default = False
Specifies if importance scores are calculated globally or per
class. Note, per class importance scores are calculated in a
one vs rest manner.
n_iter: int, default = 40
The number of iterations to run Triglav.
n_iter_fwer: int, default = 11
The iteration at which Bonferroni corrections begin.
p_1: float, default = 0.65
Used to determine the shape of the Beta-Binomial distribution
modelling hits.
p_2: float, default = 0.30
Used to determine the shape of the Beta-Binomial distribution
modelling failures.
metric: str, default = "correlation"
The dissimilarity measure used to calculate distances between
features. To use Extremely Randomized Trees proximities one
has to import 'ETCProx' from the 'triglav' package.
linkage: str, default = "complete"
The type of hierarchical clustering method to apply. The available
methods include: single, complete, ward, average, centroid.
thresh: float, default = 2.0
The threshold or max number of clusters.
criterion: str, default = "distance"
The method used to form flat clusters. The available methods
include: inconsistent, distance, maxclust, monocrit,
maxclust_monocrit.
alpha: float, default = 0.05
The level at which corrected p-values will be rejected.
run_stage_2: bool, default = True
This stage will determine the best feature from each of the
selected clusters by calculating SAGE values.
verbose: int, default = 0
Specifies if basic reporting is sent to the user.
n_jobs: int, default = 10
The number of threads
n_classes_: ndarray of shape (n_classes,)
The class labels
n_class_: int
The number of classes
selected_: ndarray of shape (n_features,)
The mask of the selected features.
selected_best_: ndarray of shape (n_features,) or None
The mask of the best features from each cluster. Only returns an ndarray
if the 'run_stage_2' parameter is enabled.
self.sage_values_: SAGE Explanation Object
A SAGE explanation object created using the set of features in 'selected_'.
For a detailed explanation on how to use this object, please visit:
https://github.com/iancovert/sage
linkage_matrix_: ndarray
The SciPy hierarchical clustering encoded as a linkage matrix.
fit(X, y, **fit_params)
Fits a `Triglav` feature selection model.
Parameters:
X: NumPy array of shape (m, n) where 'm' is the number of samples and 'n'
the number of features (taxa, OTUs, ASVs, etc).
y: NumPy array of shape (m,) where 'm' is the number of samples. Each entry
of 'y' should be a factor.
Returns:
A fitted Triglav object.
transform(X)
Reduces X to the selected features contained the 'selected_' attribute.
Input:
X: NumPy array of shape (m, n) where 'm' is the number of samples and 'n'
the number of features (taxa, OTUs, ASVs, etc).
Returns:
X_transformed: NumPy array of shape (m, p) where 'm' is the number of samples and 'p'
the number of features (taxa, OTUs, ASVs, etc). 'p' <= m
fit_transform(X, y, **fit_params)
Fits a `Triglav` feature selection model then reduces X to the selected features
contained in the 'selected_' attribute.
Parameters:
X: NumPy array of shape (m, n) where 'm' is the number of samples and 'n'
the number of features (taxa, OTUs, ASVs, etc).
y: NumPy array of shape (m,) where 'm' is the number of samples. Each entry
of 'y' should be a factor.
Returns:
X_transformed: NumPy array of shape (m, p) where 'm' is the number of samples and 'p'
the number of features (taxa, OTUs, ASVs, etc). 'p' <= m
visualize_hclust(X, y, ax, **dendrogram_kwargs)
Creates a visualization of the hierarchical clustering specified using the 'metric',
'thresh', 'linkage', and 'criterion' parameters of the `Triglav`
Parameters:
X: NumPy array of shape (m, n) where 'm' is the number of samples and 'n'
the number of features (taxa, OTUs, ASVs, etc).
y: NumPy array of shape (m,) where 'm' is the number of samples. Each entry
of 'y' should be a factor.
ax: matplotlib.axes.Axes (optional). The axes on which to plot the dendrogram. If None,
the dendrogram will be plotted to a new figure sublot axis.
Returns:
dict: A dictionary of data structures computed to render the dendrogram.
See https://docs.scipy.org/doc/scipy/reference/generated/scipy.cluster.hierarchy.dendrogram.html
for more details.
class triglav.ETCProx(n_estimators = 1024, min_samples_split = 0.33, n_sets = 5)
n_estimators: int, default = 1024
The number of estimators the ExtraTreesClassifier will use.
min_samples_split: float, default = 0.33
The minimum number of samples to split an internal node.
n_sets: int, default = 5
The number of random datasets to be generated.
fit_transform(X, y, **fit_params)
Fits a `ETCProx` class and returns the dissimilarity matrix.
Parameters:
X: NumPy array of shape (m, n) where 'm' is the number of samples and 'n'
the number of features (taxa, OTUs, ASVs, etc).
Returns:
Dissimilarity Matrix: NumPy array of shape (n, n) where 'n' is the
the number of features (taxa, OTUs, ASVs, etc)
class triglav.NoScale()
class triglav.Scaler()
class triglav.CLRTransformer()
class triglav.NoResample()
None
zero_samps: bool, ndarray of shape (n_samples,)
The mask of the all rows which sum to zeroest features from each cluster.
This is only returned for the Scaler() and CLRTransformer() classes.
fit_transform(X, y = None, **fit_params)
Fits a transformer method.
Parameters:
X: NumPy array of shape (m, n) where 'm' is the number of samples and 'n'
the number of features (taxa, OTUs, ASVs, etc).
Returns:
X_Transformed: NumPy array of shape (m, n) where 'm' is the number of
samples and 'n' the number of features (taxa, OTUs, ASVs, etc).'
NoScale will return X
Scaler will return the closure of X (all rows sum to one, X must be non-negative)
CLRTransformer will return the CLR Transform of X (X must be non-negative)
NoResample will return X