Target Contrastive Pessimistic Classifier

class libtlda.tcpr.TargetContrastivePessimisticClassifier(loss='lda', l2=1.0, max_iter=500, tolerance=1e-12, learning_rate=1.0, rate_decay='linear', verbosity=0)

Classifiers based on Target Contrastive Pessimistic Risk minimization.

Methods contain models, risk functions, parameter estimation, etc.

Methods

add_intercept(X)	Add 1’s to data as last features.
combine_class_covariances(Si, pi)	Linear combination of class covariance matrices.
discriminant_parameters(X, Y)	Estimate parameters of Gaussian distribution for discriminant analysis.
error_rate(preds, u_)	Compute classification error rate.
fit(X, y, Z)	Fit/train an importance-weighted classifier.
get_params()	Return classifier parameters.
learning_rate_t(t)	Compute current learning rate after decay.
neg_log_likelihood(X, theta)	Compute negative log-likelihood under Gaussian distributions.
predict(Z_)	Make predictions on new dataset.
predict_proba(Z)	Compute posteriors on new dataset.
project_simplex(v[, z])	Project vector onto simplex using sorting.
remove_intercept(X)	Remove 1’s from data as last features.
risk(Z, theta, q)	Compute target contrastive pessimistic risk.
tcpr_da(X, y, Z)	Target Contrastive Pessimistic Risk - discriminant analysis.

add_intercept(X)

Add 1’s to data as last features.

combine_class_covariances(Si, pi)

Linear combination of class covariance matrices.

Parameters:

Si : array

Covariance matrix (D features by D features by K classes)

pi : array

class proportions (1 by K classes)

Returns:

Si : array

Combined covariance matrix (D by D)

discriminant_parameters(X, Y)

Estimate parameters of Gaussian distribution for discriminant analysis.

Parameters:

X : array

data array (N samples by D features)

Y : array

label array (N samples by K classes)

Returns:

pi : array

class proportions (1 by K classes)

mu : array

class means (K classes by D features)

Si : array

class covariances (D features D features by K classes)

error_rate(preds, u_)

Compute classification error rate.

fit(X, y, Z)

Fit/train an importance-weighted classifier.

Parameters:

X : array

source data (N samples by D features)

y : array

source labels (N samples by 1)

Z : array

target data (M samples by D features)

Returns:

None

get_params()

Return classifier parameters.

learning_rate_t(t)

Compute current learning rate after decay.

Parameters:

t : int

current iteration

Returns:

alpha : float

current learning rate

neg_log_likelihood(X, theta)

Compute negative log-likelihood under Gaussian distributions.

Parameters:

X : array

data (N samples by D features)

theta : tuple(array, array, array)

tuple containing class proportions ‘pi’, class means ‘mu’, and class-covariances ‘Si’

Returns:

L : array

loss (N samples by K classes)

predict(Z_)

Make predictions on new dataset.

Parameters:

Z : array

new data set (M samples by D features)

Returns:

preds : array

label predictions (M samples by 1)

predict_proba(Z)

Compute posteriors on new dataset.

Parameters:

Z : array

new data set (M samples by D features)

Returns:

preds : array

label predictions (M samples by 1)

project_simplex(v, z=1.0)

Project vector onto simplex using sorting.

Reference: “Efficient Projections onto the L1-Ball for Learning in High Dimensions (Duchi, Shalev-Shwartz, Singer, Chandra, 2006).”

Parameters:

v : array

vector to be projected (n dimensions by 0)

z : float

constant (def: 1.0)

Returns:

w : array

projected vector (n dimensions by 0)

remove_intercept(X)

Remove 1’s from data as last features.

risk(Z, theta, q)

Compute target contrastive pessimistic risk.

Parameters:

Z : array

target samples (M samples by D features)

theta : array

classifier parameters (D features by K classes)

q : array

soft labels (M samples by K classes)

Returns:

float

Value of risk function.

tcpr_da(X, y, Z)

Target Contrastive Pessimistic Risk - discriminant analysis.

Parameters:

X : array

source data (N samples by D features)

y : array

source labels (N samples by 1)

Z : array

target data (M samples by D features)

Returns:

theta : array

classifier parameters (D features by K classes)