Exponentiated Gradient Versus Gradient Descent for Linear Predictors.
Jyrki KivinenManfred K. WarmuthPublished in: Inf. Comput. (1997)
Keyphrases
- exponentiated gradient
- linear predictors
- stochastic gradient
- log linear
- relative entropy
- max margin
- prediction error
- linear regression
- appearance model
- symmetric positive definite
- log linear models
- latent variables
- prior knowledge
- pixel intensities
- stochastic gradient descent
- video frames
- video data
- expectation maximization
- multi class