MuFuRU: The Multi-Function Recurrent Unit
This work addresses a bottleneck in sequence modeling for NLP tasks by enabling more flexible memory operations, though it is incremental as it builds on established RNN architectures.
The paper tackles the limitation of existing recurrent neural networks (GRU, LSTM) by proposing MuFuRUs, which allow arbitrary differentiable functions as composition operations and learn input- and state-dependent choices, resulting in improved performance in tasks like propositional logic evaluation, language modeling, and sentiment analysis.
Recurrent neural networks such as the GRU and LSTM found wide adoption in natural language processing and achieve state-of-the-art results for many tasks. These models are characterized by a memory state that can be written to and read from by applying gated composition operations to the current input and the previous state. However, they only cover a small subset of potentially useful compositions. We propose Multi-Function Recurrent Units (MuFuRUs) that allow for arbitrary differentiable functions as composition operations. Furthermore, MuFuRUs allow for an input- and state-dependent choice of these composition operations that is learned. Our experiments demonstrate that the additional functionality helps in different sequence modeling tasks, including the evaluation of propositional logic formulae, language modeling and sentiment analysis.