Joint knowledge distillation and quantization reduces AED model size to 2% of the teacher while lowering error rate by 15%.
On the compression of recurrent neural networks with an applic ation to lvcsr acoustic modeling for embedded speech recognition ,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Compression of Acoustic Event Detection Models With Quantized Distillation
Joint knowledge distillation and quantization reduces AED model size to 2% of the teacher while lowering error rate by 15%.