Row LSTM
E743714
Row LSTM is a recurrent neural network architecture used in PixelRNN that processes images row by row to model spatial dependencies for generative image modeling.
Statements (30)
| Predicate | Object |
|---|---|
| instanceOf |
component of PixelRNN
ⓘ
neural network layer ⓘ recurrent neural network architecture ⓘ |
| basedOn | Long Short-Term Memory NERFINISHED ⓘ |
| designedFor | generative image modeling ⓘ |
| domain |
computer vision
ⓘ
deep generative models ⓘ probabilistic modeling ⓘ |
| ensures | no access to future pixels in generation order ⓘ |
| hasProperty |
causal dependency structure
ⓘ
sequential row-wise computation ⓘ |
| implementedIn | PixelRNN architecture variants NERFINISHED ⓘ |
| inputType | image pixels ⓘ |
| introducedBy |
Aaron van den Oord
NERFINISHED
ⓘ
Koray Kavukcuoglu NERFINISHED ⓘ Nal Kalchbrenner NERFINISHED ⓘ |
| introducedIn | Pixel Recurrent Neural Networks NERFINISHED ⓘ |
| models | spatial dependencies in images ⓘ |
| operatesOn | 2D image grids ⓘ |
| outputType | conditional pixel distributions ⓘ |
| processes | images row by row ⓘ |
| publicationYear | 2016 ⓘ |
| publishedIn | ICML 2016 NERFINISHED ⓘ |
| relatedTo |
Diagonal BiLSTM
NERFINISHED
ⓘ
PixelCNN NERFINISHED ⓘ |
| trainingObjective | maximum likelihood estimation of pixel distributions ⓘ |
| usedFor |
autoregressive image density modeling
ⓘ
image completion ⓘ image generation ⓘ |
| usedIn | PixelRNN NERFINISHED ⓘ |
Referenced by (1)
Full triples — surface form annotated when it differs from this entity's canonical label.