straight-through estimator

E1875700 UNEXPLORED

The straight-through estimator is an approximate gradient method used to train neural networks with discrete or quantized variables by backpropagating through non-differentiable operations as if they were identity or simple differentiable functions.

Try in SPARQL Jump to: Surface forms Referenced by

All labels observed (1)

Label	Occurrences
straight-through estimator canonical	1

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.

Neural Discrete Representation Learning → optimizationTechnique → straight-through estimator ⓘ