Module liba.regress_linear
linear regression
Functions
| die (ctx) | destructor for linear regression |
| new (coef, bias) | constructor for linear regression |
| init (ctx, coef, bias) | initialize for linear regression |
| eval (ctx, val) | calculate predicted value for linear regression |
| err (ctx, x, y) | calculate residuals for linear regression |
| gd (ctx, input, error, alpha) | gradient descent for linear regression |
| sgd (ctx, x, y, alpha) | stochastic gradient descent for linear regression |
| bgd (ctx, x, y, alpha) | batch gradient descent for linear regression |
| mgd (ctx, x, y, delta, lrmax, lrmin[, lrtim[, epoch[, batch]]]) | mini-batch gradient descent for linear regression |
| zero (ctx) | zeroing for linear regression |
Functions
- die (ctx)
-
destructor for linear regression
Parameters:
- ctx a.regress_linear linear regression userdata
- new (coef, bias)
-
constructor for linear regression
Parameters:
- coef table regression coefficient
- bias number intercept
Returns:
-
a.regress_linear
linear regression userdata
- init (ctx, coef, bias)
-
initialize for linear regression
Parameters:
- ctx a.regress_linear linear regression userdata
- coef number regression coefficient
- bias number intercept
Returns:
-
a.regress_linear
linear regression userdata
- eval (ctx, val)
-
calculate predicted value for linear regression
Parameters:
- ctx a.regress_linear linear regression userdata
- val number independent variable
Returns:
-
number
predicted value
- err (ctx, x, y)
-
calculate residuals for linear regression
Parameters:
- ctx a.regress_linear linear regression userdata
- x table predictor data, specified as a numeric matrix
- y table response data, specified as a numeric vector
Returns:
-
table
residuals, specified as a numeric vector
- gd (ctx, input, error, alpha)
-
gradient descent for linear regression
Parameters:
- ctx a.regress_linear linear regression userdata
- input table predictor data, specified as a numeric vector
- error number residual, specified as a numeric scalar
- alpha number learning rate for gradient descent
Returns:
-
a.regress_linear
linear regression userdata
- sgd (ctx, x, y, alpha)
-
stochastic gradient descent for linear regression
Parameters:
- ctx a.regress_linear linear regression userdata
- x table predictor data, specified as a numeric matrix
- y table response data, specified as a numeric vector
- alpha number learning rate for gradient descent
Returns:
-
a.regress_linear
linear regression userdata
- bgd (ctx, x, y, alpha)
-
batch gradient descent for linear regression
Parameters:
- ctx a.regress_linear linear regression userdata
- x table predictor data, specified as a numeric matrix
- y table response data, specified as a numeric vector
- alpha number learning rate for gradient descent
Returns:
-
a.regress_linear
linear regression userdata
- mgd (ctx, x, y, delta, lrmax, lrmin[, lrtim[, epoch[, batch]]])
-
mini-batch gradient descent for linear regression
Parameters:
- ctx a.regress_linear linear regression userdata
- x table predictor data, specified as a numeric matrix
- y table response data, specified as a numeric vector
- delta number threshold for gradient descent value
- lrmax number maximum learning rate of iterations
- lrmin number minimum learning rate of iterations
- lrtim integer total number of learning rate steps (optional)
- epoch integer maximum number of epochs (optional)
- batch integer batch size of data (optional)
Returns:
-
number
change in loss function
- zero (ctx)
-
zeroing for linear regression
Parameters:
- ctx a.regress_linear linear regression userdata
Returns:
-
a.regress_linear
linear regression userdata