Fixed-Point RNNs: Interpolating from Diagonal to Dense
channel-wise sequence mixingcopying tasksdense linear rnnsexpressivityfixed-pointslinear recurrent neural networksparallelizable diagonal linear rnnsparameterizationssequence mixing layerssoftmax-attentionstate-of-the-art resultsstate-space modelsstate-tracking expressivitytransformer architectures
Linear recurrent neural networks (RNNs) and state-space models (SSMs) such as Mamba have become promising alternatives to softmax-attention as sequence mixing layers in Transformer architectures. Current models, however, do not exhibit the full state-tracking expressivity of RNNs because they rely on channel-wise (i.e. diagonal) sequence mixing. In this paper, we investigate parameterizations of a large class of dense linear RNNs as fixed-points of parallelizable diagonal linear RNNs. The resulting models can naturally trade expressivity for efficiency at a fixed number of parameters and achieve state-of-the-art results on the state-tracking benchmarks $A_5$ and $S_5$, while matching performance on copying and other tasks.