Parametric Models

Outliers

Specify outliers using the class type:

class_type = "outliers"

Outliers are sampled from the 2D space defined by \(x_r \times y_r\), where \(x_r\) (x range) is the closed interval of values that x can take and \(y_r\) (y range) is the closed interval of values that y can take.

"outliers": {
    "x_r": (-2.5, 2.5),
    "y_r": (-2.5, 2.5)
}

Conics

Circles

Specify circles using the class type:

class_type = "circles"

Circles are generated by uniformly sampling a center and a radius from the closed intervals specified by the user.
The default values are:

"circles": {
    "radius_r": (0.5, 1.5),
    "x_center_r": (-1.0, 1.0),
    "y_center_r": (-1.0, 1.0),
}

Lines

Specify lines using the class type:

class_type = "lines"

Lines are generated by randomly sampling two points in the 2D space \(x_r \times y_r\) to define slope and intercept, and then sampling points belonging to this line. Each point of the line is ensured to belong to the 2D space \(x_r \times y_r\).

"lines": {
    "x_r": (-2.5, 2.5),
    "y_r": (-2.5, 2.5)
}

Ellipses

Specify ellipses using the class type:

class_type = "ellipses"

Ellipses are generated by uniformly sampling a center and a radius from the closed intervals specified by the user, and then applying horizontal stretch/shrinkage to the sampled circle.
The default values are:

"ellipses": {
    "radius_r": (0.5, 1.5),
    "x_center_r": (-1, 1),
    "y_center_r": (-1, 1),
    "width_r": (0.1, 1),
    "height_r": (0.1, 1)
}

Any Conic

Specify any conic using the class type:

class_type = "conics"

This option randomly samples models from all the classes specified above. Note: Hyperbola is not implemented yet.