Fn_probabilistic_decoding

Functions
void	ecocpak::sigmoid_training (double &A, double &B, const vec f, const u32 n_pos, const u32 n_neg)
u32	ecocpak::probabilistic_decoding (const vector< Classifier * > &classifiers_vector, const vector< ClassData > &classes_vector, const imat &coding_matrix, const mat &test_samples, const ucolvec &test_set_labels, uvec &predictions, umat &confussion)
u32	probabilistic_decoding (const vector< Classifier * > &classifiers_vector, const vector< ClassData > &classes_vector, const imat &coding_matrix, const mat &test_samples, const ucolvec &test_set_labels, uvec &predictions, umat &confussion)

---

Function Documentation

u32 ecocpak::probabilistic_decoding	(	const vector< Classifier * > &	classifiers_vector,
		const vector< ClassData > &	classes_vector,
		const imat &	coding_matrix,
		const mat &	test_samples,
		const ucolvec &	test_set_labels,
		uvec &	predictions,
		umat &	confussion
	)

• Probabilistic decoding strategy.

• Input Arguments:
  • classifiers_vector : Vector of classifiers.
  • coding_matrix : Coding matrix.
  • test_set_samples : 2D matrix of test samples
  • test_set_labels : Labels vector of test samples.

• Outputs Arguments:
  • Void.

• Return Argument:
  • Integral value which denotes the number of misclassified test samples.

u32 probabilistic_decoding	(	const vector< Classifier * > &	classifiers_vector,
		const vector< ClassData > &	classes_vector,
		const imat &	coding_matrix,
		const mat &	test_samples,
		const ucolvec &	test_set_labels,
		uvec &	predictions,
		umat &	confussion
	)

• Probabilistic decoding strategy.

• Input Arguments:
  • classifiers_vector : Vector of classifiers.
  • coding_matrix : Coding matrix.
  • test_set_samples : 2D matrix of test samples
  • test_set_labels : Labels vector of test samples.

• Outputs Arguments:
  • Void.

• Return Argument:
  • Integral value which denotes the number of misclassified test samples.

void ecocpak::sigmoid_training	(	double &	A,
		double &	B,
		const vec	f,
		const u32	n_pos,
		const u32	n_neg
	)

• This is an optimization algorithm for sigmoid training. The original pseudocode of this algorithm can be found in: "Probabilistic Outputs for Support Vector Machines and Comparisons to Regularized Likehood Methods., John C. Platt., In advances in large margin classifiers., MIT Press, (1999))".

• Input Arguments:
  • f : Vector of classifiers outputs.
  • n_pos : Number of possitive samples.
  • n_neg : Number of negative samples.

• Output Arguments:
  • A : First parameter of sigmoid.
  • B : Second parameter of sigmoid.

• Return Argument:
  • Void.