Models

class Model(name)

Abstract class for models.

Parameters:

name (str) –

abstract train(*args, **kwargs)

Train the model using a given dataset

Parameters:

self (Model) –

abstract predict(data)

Predict the labels of a given set of trajectories

Parameters:

data (Data) –

Return type:

List[Any]

set_summary(**summary)

Set the summary of the model

save(path)

Save the model to a given path

Parameters:

path (str) –

evaluate(data)

Evaluate the trained model

Parameters:

data (Data) –

Return type:

Evaluation

class DecisionTreeModel(featurizer, **kwargs)

Implementation of a Decision Tree Classifier.

Parameters:

featurizer (Featurizer) –

train(data, cross_validation=0)

Train the model using a given dataset

Parameters:

  • data (Data) –

  • cross_validation (int) –

predict(data)

Predict the labels of a given set of trajectories

Parameters:

data (Data) –

Return type:

List[Any]

predict_single(traj)

Predicts the label of a single trajectory.

Parameters:

traj (Trajectory) –

Return type:

Any

class KNeighborsModel(featurizer, **kwargs)

Implementation of a K-Nearst Neighbors Classifier.

Parameters:

featurizer (Featurizer) –

train(data, cross_validation=0)

Train the model using a given dataset

Parameters:

  • data (Data) –

  • cross_validation (int) –

predict(data)

Predict the labels of a given set of trajectories

Parameters:

data (Data) –

Return type:

List[Any]

predict_single(traj)

Predicts the label of a single trajectory.

Parameters:

traj (Trajectory) –

Return type:

Any

class RandomForestModel(featurizer, **kwargs)

Implementation of a Random Forest Classifier.

Parameters:

featurizer (Featurizer) –

train(data, cross_validation=0)

Train the model using a given dataset

Parameters:

  • data (Data) –

  • cross_validation (int) –

predict(data)

Predict the labels of a given set of trajectories

Parameters:

data (Data) –

Return type:

List[Any]

predict_single(traj)

Predicts the label of a single trajectory.

Parameters:

traj (Trajectory) –

Return type:

Any

class SVMModel(featurizer, **kwargs)

Implementation of a Support Vector Machine Classifier.

Parameters:

featurizer (Featurizer) –

train(data, cross_validation=0)

Train the model using a given dataset

Parameters:

  • data (Data) –

  • cross_validation (int) –

predict(data)

Predict the labels of a given set of trajectories

Parameters:

data (Data) –

Return type:

List[Any]

predict_single(traj)

Predicts the label of a single trajectory.

Parameters:

traj (Trajectory) –

Return type:

Any

class TransformerModel(head_size=256, num_heads=1, ff_dim=4, num_transformer_blocks=2, mlp_units=None, mlp_dropout=0.4, dropout=0.25, loss='categorical_crossentropy', optimizer=None, metrics=None, max_traj_len=-1, skip_long_trajs=False, mask_value=-10000, random_state=None)

Implementation of a Transformer model.

Parameters:

  • head_size (int) –

  • num_heads (int) –

  • ff_dim (int) –

  • num_transformer_blocks (int) –

  • mlp_units (List[int] | None) –

  • mlp_dropout (float) –

  • dropout (float) –

  • max_traj_len (int) –

  • skip_long_trajs (bool) –

  • random_state (int | None) –

train(data, original_data, cross_validation=0, epochs=10, validation_split=0.2, batch_size=32, callbacks=None, checkpoint=None)

Train the model using a given dataset

Parameters:

  • data (Data) –

  • original_data (Data) –

  • cross_validation (int) –

  • epochs (int) –

  • validation_split (float) –

  • batch_size (int) –

  • callbacks (list | None) –

  • checkpoint (ModelCheckpoint | None) –

predict(data)

Predict the labels of a given set of trajectories

Parameters:

data (Data) –

Return type:

List[Any]

evaluate(data)

Evaluate the trained model

Parameters:

data (Data) –

Return type:

Evaluation

class XGBoostModel(featurizer, **kwargs)

Implementation of a XGBoost Classifier.

Parameters:

featurizer (Featurizer) –

train(data, cross_validation=0)

Train the model using a given dataset

Parameters:

  • data (Data) –

  • cross_validation (int) –

predict(data)

Predict the labels of a given set of trajectories

Parameters:

data (Data) –

Return type:

List[Any]

predict_single(traj)

Predicts the label of a single trajectory.

Parameters:

traj (Trajectory) –

Return type:

Any

class LSTMModel(units=None, masking_value=None, loss='sparse_categorical_crossentropy', optimizer='rmsprop', metrics=None, random_state=None, **kwargs)

Implementation of a LSTM Classifier.

Parameters:

  • units (List[int] | None) –

  • masking_value (int | None) –

  • random_state (int | None) –

train(data, dataset, cross_validation=0, epochs=10, batch_size=None, validation_split=None, callbacks=None, checkpoint=None)

Train the model using a given dataset

Parameters:

  • data (Data) –

  • dataset (Dataset) –

  • callbacks (list | None) –

  • checkpoint (ModelCheckpoint | None) –

predict(data)

Predict the labels of a given set of trajectories

Parameters:

data (Data) –

Return type:

List[Any]

evaluate(data)

Evaluate the trained model

Parameters:

data (Data) –

Return type:

Evaluation