Statements (53)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
gptkb:artificial_intelligence
|
| gptkbp:abbreviation |
gptkb:RNN
|
| gptkbp:activatedBy |
gptkb:ReLU
sigmoid tanh |
| gptkbp:canBeRegularizedBy |
dropout
weight decay |
| gptkbp:firstDescribed |
1980s
|
| gptkbp:hasApplication |
text generation
sentiment analysis chatbots |
| gptkbp:hasComponent |
hidden state
input layer output layer recurrent connections |
| gptkbp:hasVariant |
gptkb:Bidirectional_RNN
gptkb:Echo_State_Network Gated Recurrent Unit Long Short-Term Memory |
| https://www.w3.org/2000/01/rdf-schema#label |
Recurrent neural networks
|
| gptkbp:implementedIn |
gptkb:TensorFlow
gptkb:Keras gptkb:Theano gptkb:PyTorch |
| gptkbp:improves |
attention mechanism
|
| gptkbp:input |
sequential data
|
| gptkbp:inventedBy |
gptkb:Geoffrey_Hinton
gptkb:John_Hopfield gptkb:David_Rumelhart gptkb:Ronald_J._Williams |
| gptkbp:limitation |
difficulty learning long-term dependencies
|
| gptkbp:numberOfIssues |
exploding gradient problem
vanishing gradient problem |
| gptkbp:output |
sequential data
single value |
| gptkbp:relatedTo |
convolutional neural networks
transformers feedforward neural networks |
| gptkbp:stackable |
deep RNN
|
| gptkbp:trainer |
gptkb:backpropagation_through_time
|
| gptkbp:usedFor |
machine translation
natural language processing speech recognition time series prediction handwriting recognition sequence modeling |
| gptkbp:usedIn |
robotics
financial forecasting language modeling music generation video analysis |
| gptkbp:bfsParent |
gptkb:Neural_Engine
|
| gptkbp:bfsLayer |
6
|