From on-line prediction wiki

Main: LeastSquares

Least squares method for online linear regression is a method which finds the best fitted predictor for the past data, and then predicts following this predictor. It finds the argument of $\min_\theta \left(\sum_{t=1}^T (\theta' x_t - y_t)^2 \right)$ after $T$ steps.

See more http://en.wikipedia.org/wiki/Least_squares.

Retrieved from ?n=Main.LeastSquares
Page last modified on August 23, 2009, at 05:57 PM