Approximating Stacked and Bidirectional Recurrent Architectures with the Delayed Recurrent Neural Network
This work addresses efficiency and expressiveness challenges in sequence modeling for researchers and practitioners, offering a novel architectural simplification that is incremental but potentially impactful.
The paper tackles the problem of enhancing recurrent neural networks (RNNs) by proposing the delayed-RNN, a single-layer architecture with a delay between input and output, and shows it can approximate stacked and bidirectional RNNs with similar or faster runtimes, achieving comparable performance in synthetic and real-world NLP tasks.
Recent work has shown that topological enhancements to recurrent neural networks (RNNs) can increase their expressiveness and representational capacity. Two popular enhancements are stacked RNNs, which increases the capacity for learning non-linear functions, and bidirectional processing, which exploits acausal information in a sequence. In this work, we explore the delayed-RNN, which is a single-layer RNN that has a delay between the input and output. We prove that a weight-constrained version of the delayed-RNN is equivalent to a stacked-RNN. We also show that the delay gives rise to partial acausality, much like bidirectional networks. Synthetic experiments confirm that the delayed-RNN can mimic bidirectional networks, solving some acausal tasks similarly, and outperforming them in others. Moreover, we show similar performance to bidirectional networks in a real-world natural language processing task. These results suggest that delayed-RNNs can approximate topologies including stacked RNNs, bidirectional RNNs, and stacked bidirectional RNNs - but with equivalent or faster runtimes for the delayed-RNNs.