SLP¶
This module implements the algorithm SLP.
References
| [1] | D.-M. Liang and Y.-F. Li. Lightweight Label Propagation for Large-Scale Network Data. In: 27th International Joint Conference on Artificial Intelligence (IJCAI‘18), Stockholm, Sweden, 2018. |
- Author:
- De-Ming Liang <XXX@gmail.com> Xiao-Shuang Lv <XXX@XXX.com>
- License:
- MIT
-
class
s3l.data_quality.SLP.SLP(stepSize=0.1, T=6)[source]¶ Bases:
s3l.base.TransductiveEstimatorwithGraphThis is a python implementation of SLP, which can do label propagation on large-scale graphs.
Read more in the User Guide.
Parameters: - stepSize (coefficient, optical (default=0.1)) – step size.
- T (coefficient,optical (default=6)) – running epoches.
Examples
>>> from s3l.data_quality.SLP import SLP >>> from s3l.datasets import data_manipulate, base >>> X, y = base.load_covtype(True) >>> W = base.load_graph_covtype(True) >>> _, test_idxs, labeled_idxs, unlabeled_idxs = \ >>> data_manipulate.inductive_split(X=X, y=y) >>> slp = SLP(stepSize=0.1, T=6) >>> slp.fit(X,y,labeled_idxs,W) >>> slp.predict(unlabeled_idxs) [1,-1,-1,1,1...,1]
References
SLP implements the LEAD algorithm in [1].
[1] D.-M. Liang and Y.-F. Li. Lightweight Label Propagation for Large-Scale Network Data. In: 27th International Joint Conference on Artificial Intelligence (IJCAI‘18), Stockholm, Sweden, 2018. -
fit(X, y, l_ind, W)[source]¶ Fit the model to data.
Parameters: - W (sparse matrix) – affinity matrix, labels should be at the left-top corner, should be in sparse form.
- y (array-like) – label vector with different labels [n_samples].
- l_ind (array-like) – a row vector with length l, where l is the number of labeled instance. Each element is an index of a labeled instance.
Returns: pred – prediction of labels [n_samples, n_labels], in the original sort.
Return type: array-like
-
predict(u_ind)[source]¶ Compute the most possible label for samples in W.
Parameters: u_ind (array-like) – a row vector with length l, where l is the number of unlabeled instance. Each element is an index of a unlabeled instance. Returns: pred – Each row is the most likely label for a sample [n_samples]. Return type: array-like
-
predict_proba(u_ind)[source]¶ Compute probabilities of possible labels for samples in W.
Parameters: u_ind (array-like) – a row vector with length l, where l is the number of unlabeled instance. Each element is an index of a unlabeled instance. Returns: pred – Each line is the probability of possible labels of a sample involved in the calculation of the prediction [n_samples, n_labels]. Return type: array-like