Skip to main content
formatting
Source Link
amoeba
  • 107.2k
  • 36
  • 321
  • 346

I understand that the ridge regression estimate is the $\beta$ that minimizes residual sum of square and a penalty on the size of $\beta$

$$\beta_{ridge} = (\lambda I_D + X'X)^{-1}X'y = \text{argmin}\: RSS + \lambda ||\beta||^2_2$$$$\beta_\mathrm{ridge} = (\lambda I_D + X'X)^{-1}X'y = \operatorname{argmin}\big[ \text{RSS} + \lambda \|\beta\|^2_2\big]$$

However, I don't fully understand the significance of the fact that $\beta_{ridge}$$\beta_\text{ridge}$ differs from $\beta_{OLS}$$\beta_\text{OLS}$ by only adding a small constant to the diagonal of $X'X$. Indeed,

$$\beta_{OLS} = (X'X)^{-1}X'y$$$$\beta_\text{OLS} = (X'X)^{-1}X'y$$

  1. My book mentions that this makes the estimate more stable numerically -- why?

  2. Is numerical stability related to the shrinkage towards 0 of the ridge estimate, or it's just a coincidence?

I understand that the ridge regression estimate is the $\beta$ that minimizes residual sum of square and a penalty on the size of $\beta$

$$\beta_{ridge} = (\lambda I_D + X'X)^{-1}X'y = \text{argmin}\: RSS + \lambda ||\beta||^2_2$$

However, I don't fully understand the significance of the fact that $\beta_{ridge}$ differs from $\beta_{OLS}$ by only adding a small constant to the diagonal of $X'X$. Indeed,

$$\beta_{OLS} = (X'X)^{-1}X'y$$

  1. My book mentions that this makes the estimate more stable numerically -- why?

  2. Is numerical stability related to the shrinkage towards 0 of the ridge estimate, or it's just a coincidence?

I understand that the ridge regression estimate is the $\beta$ that minimizes residual sum of square and a penalty on the size of $\beta$

$$\beta_\mathrm{ridge} = (\lambda I_D + X'X)^{-1}X'y = \operatorname{argmin}\big[ \text{RSS} + \lambda \|\beta\|^2_2\big]$$

However, I don't fully understand the significance of the fact that $\beta_\text{ridge}$ differs from $\beta_\text{OLS}$ by only adding a small constant to the diagonal of $X'X$. Indeed,

$$\beta_\text{OLS} = (X'X)^{-1}X'y$$

  1. My book mentions that this makes the estimate more stable numerically -- why?

  2. Is numerical stability related to the shrinkage towards 0 of the ridge estimate, or it's just a coincidence?

Tweeted twitter.com/#!/StackStats/status/521195687061762048
formatting
Source Link
Glen_b
  • 290.4k
  • 37
  • 652
  • 1.1k

I understand that the ridge regression estimate is the $\beta$ that minimizes residual sum of square and a penalty on the size of $\beta$

$$\beta_{ridge} = (\lambda I_D + X'X)^{-1}X'y = argmin RSS + \lambda ||\beta||^2_2$$$$\beta_{ridge} = (\lambda I_D + X'X)^{-1}X'y = \text{argmin}\: RSS + \lambda ||\beta||^2_2$$

However, I don't fully understand the significance of the fact that $\beta_{ridge}$ differs from $\beta_{OLS}$ by only adding a small constant to the diagonal of $X'X$. Indeed,

$$\beta_{OLS} = (X'X)^{-1}X'y$$

  1. My book mentions that this makes the estimate more stable numerically -- why?

  2. Is numerical stability related to the shrinkage towards 0 of the ridge estimate, or it's just a coincidence?

I understand that the ridge regression estimate is the $\beta$ that minimizes residual sum of square and a penalty on the size of $\beta$

$$\beta_{ridge} = (\lambda I_D + X'X)^{-1}X'y = argmin RSS + \lambda ||\beta||^2_2$$

However, I don't fully understand the significance of the fact that $\beta_{ridge}$ differs from $\beta_{OLS}$ by only adding a small constant to the diagonal of $X'X$. Indeed,

$$\beta_{OLS} = (X'X)^{-1}X'y$$

  1. My book mentions that this makes the estimate more stable numerically -- why?

  2. Is numerical stability related to the shrinkage towards 0 of the ridge estimate, or it's just a coincidence?

I understand that the ridge regression estimate is the $\beta$ that minimizes residual sum of square and a penalty on the size of $\beta$

$$\beta_{ridge} = (\lambda I_D + X'X)^{-1}X'y = \text{argmin}\: RSS + \lambda ||\beta||^2_2$$

However, I don't fully understand the significance of the fact that $\beta_{ridge}$ differs from $\beta_{OLS}$ by only adding a small constant to the diagonal of $X'X$. Indeed,

$$\beta_{OLS} = (X'X)^{-1}X'y$$

  1. My book mentions that this makes the estimate more stable numerically -- why?

  2. Is numerical stability related to the shrinkage towards 0 of the ridge estimate, or it's just a coincidence?

added 11 characters in body
Source Link
Heisenberg
  • 4.6k
  • 4
  • 31
  • 65

I understand that the ridge regression estimate is the $\beta$ that minimizes residual sum of square and a penalty on the size of $\beta$

$$\beta_{ridge} = (\lambda I_D + X'X)^{-1}X'y = argmin RSS + \lambda ||\beta||^2_2$$

However, I don't fully understand the significance of the fact that $\beta_{ridge}$ differs from $\beta_{OLS}$ by only adding a small constant to the diagonal of $X'X$. Indeed,

$$\beta_{OLS} = (X'X)^{-1}X'y$$

  1. My book mentions that this makes the estimate more stable numerically -- why?

  2. Is numerical stability related to the sparsityshrinkage towards 0 of the ridge estimate, or it's just a coincidence?

I understand that the ridge regression estimate is the $\beta$ that minimizes residual sum of square and a penalty on the size of $\beta$

$$\beta_{ridge} = (\lambda I_D + X'X)^{-1}X'y = argmin RSS + \lambda ||\beta||^2_2$$

However, I don't fully understand the significance of the fact that $\beta_{ridge}$ differs from $\beta_{OLS}$ by only adding a small constant to the diagonal of $X'X$. Indeed,

$$\beta_{OLS} = (X'X)^{-1}X'y$$

  1. My book mentions that this makes the estimate more stable numerically -- why?

  2. Is numerical stability related to the sparsity of the ridge estimate, or it's just a coincidence?

I understand that the ridge regression estimate is the $\beta$ that minimizes residual sum of square and a penalty on the size of $\beta$

$$\beta_{ridge} = (\lambda I_D + X'X)^{-1}X'y = argmin RSS + \lambda ||\beta||^2_2$$

However, I don't fully understand the significance of the fact that $\beta_{ridge}$ differs from $\beta_{OLS}$ by only adding a small constant to the diagonal of $X'X$. Indeed,

$$\beta_{OLS} = (X'X)^{-1}X'y$$

  1. My book mentions that this makes the estimate more stable numerically -- why?

  2. Is numerical stability related to the shrinkage towards 0 of the ridge estimate, or it's just a coincidence?

added 42 characters in body; edited tags
Source Link
Heisenberg
  • 4.6k
  • 4
  • 31
  • 65
Loading
Source Link
Heisenberg
  • 4.6k
  • 4
  • 31
  • 65
Loading