Statements (23)
Predicate | Object |
---|---|
gptkbp:instanceOf |
gptkb:algorithm
|
gptkbp:appliesTo |
natural language processing
speech recognition time series prediction |
gptkbp:approach |
splits long sequences into shorter segments
|
gptkbp:category |
gptkb:machine_learning
deep learning |
gptkbp:firstDescribed |
1990s
|
gptkbp:fullName |
gptkb:Truncated_Backpropagation_Through_Time
|
https://www.w3.org/2000/01/rdf-schema#label |
Truncated BPTT
|
gptkbp:input |
sequence data
|
gptkbp:limitation |
may miss long-term dependencies
|
gptkbp:output |
updated neural network weights
|
gptkbp:reduces |
computational cost
memory usage |
gptkbp:relatedTo |
gradient descent
Backpropagation Through Time RNN training |
gptkbp:solvedBy |
computational inefficiency of full BPTT
|
gptkbp:tradeoff |
approximate gradients
|
gptkbp:usedFor |
training recurrent neural networks
|
gptkbp:bfsParent |
gptkb:BPTT
|
gptkbp:bfsLayer |
8
|