svr_linear – Support vector regression with linear kernel.#

Back to family axis | Back to L4 | Browse all options

Operational op under axis family, sub-layer L4_A_model_selection, layer l4. Standalone callable: mf.functions.svr_linear_fit.

Function signature#

mf.functions.svr_linear_fit(
    X: np.ndarray | pd.DataFrame,
    y: np.ndarray | pd.Series,
) -> SVRLinearFitResult

Parameters#

name

type

default

constraint

description

X

`np.ndarray

pd.DataFrame`

—

—

y

`np.ndarray

pd.Series`

—

—

Returns#

SVRLinearFitResult — frozen dataclass with fit results.

Attribute

Type

Description

.C

float

Regularisation parameter used.

.n_support_vectors

int

Number of support vectors.

.predict(X)

np.ndarray

Predictions for new data X, shape (n_samples,).

.summary()

str

Table: C and support vector count.

Behavior#

ε-insensitive loss + L2 regularisation. Sparse in support vectors.

Configures the family axis on L4_A_model_selection (layer l4); the svr_linear value is materialised in the recipe’s fixed_axes block under that sub-layer.

When to use

Robust linear baselines; comparison against ridge.

In recipe context#

Set params.family = "svr_linear" in the relevant layer to activate this op within a recipe:

# Layer L4 recipe fragment
params:
  family: svr_linear

References#

  • macroforecast design Part 2, L4: ‘forecasting model is the layer where every authoring iteration ends – pick family, tune, repeat.’

  • Drucker, Burges, Kaufman, Smola & Vapnik (1997) ‘Support Vector Regression Machines’, NeurIPS.