W0 in Linear Discriminant analyis

[rahls7]

When you take the threshold while predicting y, you did not account for the w0 terms in your equation. Hence, it always predicts y as 1, because h is always negative. The equation should be something like this.

[CodeName-Detective]

@rahls7 Very much True. Not only that. The estimation of the covariance matrix should be weighted average of the covariance matrix of all the classes. But here he considered the sum of covariance matrix as estimation.