EBMs Trained with Maximum Likelihood are Generator Models Trained with a Self-adverserial Loss.
Zhisheng XiaoQing YanYali AmitPublished in: CoRR (2021)
Keyphrases
- maximum likelihood
- experimental data
- training set
- statistical models
- multi layered perceptron
- probabilistic model
- mathematical models
- model selection
- multilayer perceptron
- svm classifier
- data sets
- discriminative training
- accurate models
- training process
- parameter estimation
- complex systems
- process model
- speech recognition
- support vector machine svm
- em algorithm
- information systems
- machine learning
- neural network
- trained classifiers
- real time