|
gptkbp:instanceOf
|
gptkb:convolutional_neural_network
energy-based model
probabilistic graphical model
|
|
gptkbp:activatedBy
|
sigmoid
sigmoid function
|
|
gptkbp:alsoKnownAs
|
gptkb:Bayes_network
gptkb:Bayesian_belief_network
gptkb:belief_network
|
|
gptkbp:alternativeName
|
Autoencoder
Bayesian_network
Restricted_Boltzmann_Machine
|
|
gptkbp:application
|
gptkb:gene_regulatory_networks
computer vision
information retrieval
robotics
speech recognition
bioinformatics
risk analysis
diagnosis
|
|
gptkbp:canBe
|
optimization
pattern recognition
data representation
probabilistic inference
learned from data
specified by experts
|
|
gptkbp:component
|
edges represent conditional dependencies
nodes represent random variables
|
|
gptkbp:energyFunction
|
gptkb:Boltzmann_distribution
E(v,h) = -a^Tv - b^Th - v^TWh
|
|
gptkbp:extendsTo
|
gptkb:Deep_Boltzmann_machine
gptkb:Replicated_Softmax_model
gptkb:Restricted_Boltzmann_machine
|
|
gptkbp:field
|
gptkb:artificial_intelligence
gptkb:machine_learning
computational neuroscience
deep learning
Machine learning
Deep learning
|
|
gptkbp:generalizes
|
gptkb:Hopfield_network
|
|
gptkbp:hasApplication
|
dimensionality reduction
feature extraction
collaborative filtering
image modeling
speech modeling
|
|
gptkbp:hasComponent
|
gptkb:Decoder
hidden units
stochastic binary units
visible units
Encoder
|
|
gptkbp:hasGraphStructure
|
fully connected
|
|
gptkbp:hasLayerType
|
hidden layer
visible layer
|
|
gptkbp:hasLearningRule
|
gptkb:Gibbs_sampling
gptkb:Hebbian_learning
|
|
gptkbp:hasModel
|
joint probability distribution
|
|
gptkbp:hasNoDirectConnectionsBetween
|
hidden units
visible units
|
|
gptkbp:hasNoIntraLayerConnections
|
true
|
|
gptkbp:hasVariant
|
gptkb:Conditional_RBM
gptkb:Gaussian-Bernoulli_RBM
gptkb:Multinomial_RBM
gptkb:Replicated_Softmax_RBM
|
|
gptkbp:hiddenUnits
|
h
|
|
gptkbp:includesState
|
binary
|
|
gptkbp:input
|
gptkb:Vector
binary
|
|
gptkbp:inspiredBy
|
gptkb:statistical_mechanics
gptkb:Deep_Belief_Network
|
|
gptkbp:introduced
|
gptkb:Geoffrey_Hinton
gptkb:Terry_Sejnowski
|
|
gptkbp:introducedIn
|
1980s
1985
1986
|
|
gptkbp:inventedBy
|
gptkb:Geoffrey_Hinton
|
|
gptkbp:isBipartite
|
true
|
|
gptkbp:isGenerative
|
true
|
|
gptkbp:isStochastic
|
true
|
|
gptkbp:isUndirected
|
true
|
|
gptkbp:limitation
|
computationally expensive
difficult to scale
slow training
cannot represent cyclic dependencies
complexity increases with number of variables
structure learning is NP-hard
parameter estimation can be difficult with sparse data
cannot model sequential data directly
difficult to train for large networks
|
|
gptkbp:namedAfter
|
gptkb:Ludwig_Boltzmann
|
|
gptkbp:optimizedFor
|
stochastic gradient descent
|
|
gptkbp:originatedIn
|
gptkb:Judea_Pearl
|
|
gptkbp:output
|
gptkb:Vector
binary
|
|
gptkbp:parameter
|
weights
biases
|
|
gptkbp:probabilisticModel
|
gptkb:Markov_Random_Field
|
|
gptkbp:property
|
encodes joint probability distribution
supports causal discovery
supports efficient inference algorithms
supports learning from incomplete data
supports parameter learning
supports structure learning
|
|
gptkbp:reduces
|
Reconstruction error
|
|
gptkbp:relatedTo
|
gptkb:Principal_component_analysis
gptkb:artificial_intelligence
gptkb:probability_theory
gptkb:Markov_network
gptkb:Boltzmann_Machine
gptkb:Deep_Belief_Network
gptkb:Hopfield_network
gptkb:Restricted_Boltzmann_machine
Boltzmann machine
statistics
graph theory
causal inference
Neural network
|
|
gptkbp:represents
|
set of variables and their conditional dependencies
|
|
gptkbp:samplingMethod
|
gptkb:Gibbs_sampling
|
|
gptkbp:stackable
|
true
|
|
gptkbp:supportsAlgorithm
|
gptkb:Markov_Chain_Monte_Carlo
gptkb:Gibbs_sampling
belief propagation
expectation-maximization
variable elimination
|
|
gptkbp:trainer
|
Unsupervised learning
|
|
gptkbp:trainingAlgorithm
|
stochastic gradient descent
contrastive divergence
|
|
gptkbp:usedFor
|
gptkb:dictionary
gptkb:machine_learning
decision making
diagnosis
dimensionality reduction
unsupervised learning
collaborative filtering
feature learning
generative modeling
probabilistic inference
prediction
Data compression
causal reasoning
Dimensionality reduction
Denoising
Anomaly detection
Feature learning
|
|
gptkbp:usedIn
|
gptkb:Netflix_Prize
unsupervised learning
pretraining deep networks
|
|
gptkbp:uses
|
gptkb:directed_acyclic_graph
|
|
gptkbp:variant
|
gptkb:Contractive_autoencoder
gptkb:Denoising_autoencoder
gptkb:Sparse_autoencoder
gptkb:Variational_autoencoder
|
|
gptkbp:visibleUnits
|
v
|
|
gptkbp:bfsParent
|
gptkb:Geoffrey_Hinton
|
|
gptkbp:bfsLayer
|
4
|