Module probability 

Probability estimation functions for SVM models.

Provides Platt scaling (sigmoid_train/sigmoid_predict), multiclass probability estimation, and density-based probability for one-class SVM. Matches the original LIBSVM’s probability routines (svm.cpp:1714–2096).

Functions

multiclass_probability

Solve multiclass probabilities from pairwise estimates.

predict_one_class_probability

Predict probability for one-class SVM from density marks.

sigmoid_predict

Numerically stable sigmoid prediction.

sigmoid_train

Train Platt scaling parameters (A, B) via Newton’s method.

svm_binary_svc_probability

Estimate Platt scaling parameters for a binary sub-problem.

svm_one_class_probability

Estimate probability density marks for one-class SVM.

svm_svr_probability

Estimate Laplace scale parameter for SVR probability.