gbnet.gblinear
==============

.. py:module:: gbnet.gblinear


Classes
-------

.. autoapisummary::

   gbnet.gblinear.GBLinear


Functions
---------

.. autoapisummary::

   gbnet.gblinear.ridge_regression


Module Contents
---------------

.. py:class:: GBLinear(input_dim, output_dim, bias = True, lr = 0.5, min_hess = 0.0, lambd = 0.01)

   Bases: :py:obj:`gbnet.base.BaseGBModule`


   A linear gradient boosting module that uses gradient boosting for updates.

   This module implements a linear layer that can be trained using gradient boosting.
   It maintains state between iterations and updates parameters using computed gradients
   and hessians.

   :param input_dim: Input feature dimension
   :type input_dim: int
   :param output_dim: Output prediction dimension
   :type output_dim: int
   :param bias: Whether to include a bias term. Defaults to True.
   :type bias: bool, optional
   :param lr: Learning rate for parameter updates. Defaults to 0.5.
   :type lr: float, optional
   :param min_hess: Minimum hessian threshold. Defaults to 0.0.
   :type min_hess: float, optional
   :param lambd: L2 regularization parameter. Defaults to 0.01.
   :type lambd: float, optional

   :ivar linear: The underlying linear layer
   :vartype linear: nn.Linear
   :ivar FX: Current predictions tensor
   :vartype FX: torch.Tensor
   :ivar input: Input data cache
   :vartype input: numpy.ndarray
   :ivar g: Gradient cache
   :vartype g: torch.Tensor
   :ivar h: Hessian cache

   :vartype h: torch.Tensor


   .. py:attribute:: input_dim


   .. py:attribute:: output_dim


   .. py:attribute:: min_hess
      :value: 0.0



   .. py:attribute:: bias
      :value: True



   .. py:attribute:: lr
      :value: 0.5



   .. py:attribute:: lambd
      :value: 0.01



   .. py:attribute:: linear


   .. py:attribute:: FX
      :value: None



   .. py:attribute:: input
      :value: None



   .. py:attribute:: g
      :value: None



   .. py:attribute:: h
      :value: None



   .. py:method:: _input_checking_setting(x)


   .. py:method:: forward(x)


   .. py:method:: gb_calc()

      Calculate gradients and stores in the object



   .. py:method:: gb_step()

      Uses stored gradients to update weights



.. py:function:: ridge_regression(X, y, lambd)

   Solves ridge regression using Cholesky decomposition.

   Fastest method tested.

   :param X: Design matrix of shape (n_samples, n_features)
   :type X: np.ndarray
   :param y: Target values of shape (n_samples,)
   :type y: np.ndarray
   :param lambd: Ridge regularization parameter
   :type lambd: float

   :returns: Fitted coefficients of shape (n_features,)
   :rtype: np.ndarray

   The function solves the ridge regression problem:
   min_beta ||X beta - y||^2 + lambd ||beta||^2
   using the normal equations and Cholesky decomposition for numerical stability.