lal.core package

Submodules

lal.core.model module

This is a combination of the weighting methodology and the matching methodology.

The feature weights are used to scale the training and testing features, and then used to help the matching algorithm which samples are closest to each other based on which one is more important to predict a particular value.

class lal.core.model.LALGBClassifier(**kwargs)

Bases: lal.core.model._LALGBBaseModel

This is when our training labels are categorical.

predict(**kwargs)

We choose most probable label our samples in the testing dataset has.

Parameters

• train_data (numpy.array) – The training dataset features

• train_labels (numpy.array) – The training dataset labels

• test_data (numpy.array) – The testing dataset features

Returns

predict_proba(**kwargs)

This predicts the probability of our test data having any of the available labels in the training dataset

Parameters

• train_data (numpy.array) – The training dataset features

• train_labels (numpy.array) – The training dataset labels

• test_data (numpy.array) – The testing dataset features

Returns

class lal.core.model.LALGBRegressor(**kwargs)

Bases: lal.core.model._LALGBBaseModel

This is when our training labels are continuous.

predict(**kwargs)

We predict the possible value our testing dataset will have, based on the continuous variables.

Parameters

• train_data (numpy.array) – The training dataset features

• train_labels (numpy.array) – The training dataset labels

• test_data (numpy.array) – The testing dataset features

Returns

lal.core.nn module

This matching methodology and the distance measure calculations exist here. Please consider the remaining code base to learn more about which algorithms are available.

class lal.core.nn.EMDCosineMatcher(training_weights, testing_weights, thrsh)

Bases: lal.core.nn._CosineDistance, lal.core.nn._EMDMatcher

This is the Earth Mover Distance Matching algorithm with the cosine distance measure.

class lal.core.nn.EMDMahalanobisMatcher(training_weights, testing_weights, thrsh)

Bases: lal.core.nn._MahalanobisDistance, lal.core.nn._EMDMatcher

This is the Earth Mover Distance Matching algorithm with the mahalanobis distance measure.

class lal.core.nn.EMDPowerMatcher(p, training_weights, testing_weights, thrsh)

Bases: lal.core.nn._PowerDistance, lal.core.nn._EMDMatcher

This is the Earth Mover Distance Matching algorithm with the p-norm distance measure.

class lal.core.nn.KNNCosineMatcher(k)

Bases: lal.core.nn._CosineDistance, lal.core.nn._KNNBase

This is the K-Nearest Neighbor algorithm with the cosine distance measure.

class lal.core.nn.KNNMahalanobisMatcher(k)

Bases: lal.core.nn._MahalanobisDistance, lal.core.nn._KNNBase

This is the K-Nearest Neighbor algorithm with the mahalanobis distance measure.

class lal.core.nn.KNNPowerMatcher(k, p)

Bases: lal.core.nn._PowerDistance, lal.core.nn._KNNBase

This is the K-Nearest Neighbor algorithm with the p-norm distance measure.

class lal.core.nn.NNLinearSumCosineMatcher

Bases: lal.core.nn._CosineDistance, lal.core.nn._NNLinearSumBase

This is the Exhaustive-Hungarian Matching algorithm with the cosine distance measure.

class lal.core.nn.NNLinearSumMahalanobisMatcher

Bases: lal.core.nn._MahalanobisDistance, lal.core.nn._NNLinearSumBase

This is the Exhaustive-Hungarian Matching algorithm with the mahalanobis distance measure.

class lal.core.nn.NNLinearSumPowerMatcher(p)

Bases: lal.core.nn._PowerDistance, lal.core.nn._NNLinearSumBase

This is the Exhaustive-Hungarian Matching algorithm with the p-norm distance measure.

lal.core.weights module

Given the type of regression task, we decide the optimal parameters for the base learner, by default is the Xgboost mode, and then use the feature importance to devise feature weights for each feature in the training set.

class lal.core.weights.GBMClassifierWeight

Bases: lal.core.weights._GBMWeightsBase

This is for our classification-task

class lal.core.weights.GBMRegressorWeight

Bases: lal.core.weights._GBMWeightsBase

This is for our regression-task.

Module contents