MIT Press

Skip to Main Content

Table 1:

Algorithmic Test on GORU, GRU, LSTM, EURNN, oRNN, and uRNN.

Model	Accuracy
LSTM	$70.3 \pm 1.1$
GRU	$64.2 \pm 2.1$
uRNN	$66.3 \pm 2.2$
oRNN	$44.5 \pm 2.4$
EURNN	$54.5 \pm 0.9$
GORU (EURNN FFT-style)	$73.9 \pm 1.4$
GORU (Householder)	$73.6 \pm 2.3$

Notes: GORU significantly outperforms all other RNN models. The bold number represents the highest accuracy.