A framework unifies recent online RNN training algorithms along four axes and demonstrates performance clustering on synthetic tasks, indicating that gradient alignment is insufficient to explain success especially for stochastic methods.
On the Variance of Unbiased Online Recurrent Optimization
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
The recently proposed Unbiased Online Recurrent Optimization algorithm (UORO, arXiv:1702.05043) uses an unbiased approximation of RTRL to achieve fully online gradient-based learning in RNNs. In this work we analyze the variance of the gradient estimate computed by UORO, and propose several possible changes to the method which reduce this variance both in theory and practice. We also contribute significantly to the theoretical and intuitive understanding of UORO (and its existing variance reduction technique), and demonstrate a fundamental connection between its gradient estimate and the one that would be computed by REINFORCE if small amounts of noise were added to the RNN's hidden units.
fields
cs.LG 1years
2019 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
A Unified Framework of Online Learning Algorithms for Training Recurrent Neural Networks
A framework unifies recent online RNN training algorithms along four axes and demonstrates performance clustering on synthetic tasks, indicating that gradient alignment is insufficient to explain success especially for stochastic methods.