Additive versus exponentiated gradient updates for linear prediction.
Jyrki KivinenManfred K. WarmuthPublished in: STOC (1995)
Keyphrases
- linear prediction
- exponentiated gradient
- image coding
- stochastic gradient
- log linear
- max margin
- lossless compression
- prediction error
- relative entropy
- speech signal
- subband
- autoregressive
- symmetric positive definite
- machine learning
- multi class
- spectral analysis
- image processing
- multimedia
- log linear models
- computational complexity
- hidden markov models
- image compression
- latent variables
- non stationary
- filter bank