Fundamental operating regimes, hyper-parameter fine-tuning and glassiness: towards an interpretable replica-theory for trained restricted Boltzmann machines.
Alberto FachechiElena AgliariMiriam AquaroAnthony CoolenMenno MulderPublished in: CoRR (2024)
Keyphrases
- fine tuning
- contrastive divergence
- restricted boltzmann machine
- hyperparameters
- optical flow
- deep learning
- closed form
- model selection
- random sampling
- parameter settings
- hidden layer
- bayesian framework
- em algorithm
- prior information
- noise level
- multilayer perceptron
- data mining
- cross validation
- sample size
- maximum likelihood
- semi supervised
- prior knowledge
- feature space