paper-with-me

Papers

Improving the Performance and Learning Stability of Parallelizable RNNs Designed for Ultra-Low Power Applications

2026-05-12 · Julien Brandoit, Arthur Fyon, Damien Ernst, Guillaume Drion arxiv

Sequence learning is dominated by Transformers and parallelizable recurrent neural networks (RNNs) such as state-space models, yet learning long-term dependencies remains challenging, and state-of-the-art designs trade power consumption for performance. The Bistable Memory Recurrent Unit (BMRU) was introduced to enable hardware-software co-design of ultra-low power RNNs: quantized states with hysteresis provide persistent memory while mapping directly to analog primitives. However, BMRU performance lags behind parallelizable RNNs on complex sequential tasks. In this paper, we identify gradient blocking during state updates as a key limitation and propose a cumulative update formulation that restores gradient flow while preserving persistent memory, creating skip-connections through time. This leads to the Cumulative Memory Recurrent Unit (CMRU) and its relaxed variant, the $α$CMRU. Experiments show that the cumulative formulation dramatically improves convergence stability and reduces initialization sensitivity. The CMRU and $α$CMRU match or outperform Linear Recurrent Units (LRUs) and minimal Gated Recurrent Units (minGRUs) across diverse benchmarks at small model sizes, with particular advantages on tasks requiring discrete long-range retention, while the CMRU retains quantized states, persistent memory, and noise-resilient dynamics essential for analog implementation.

📄 PDF Abstract BibTeX arXiv:2605.11855

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Parallelizable memory recurrent units

2026-01-14 · Florent De Geeter, Gaspard Lambrechts, Damien Ernst, Guillaume Drion arxiv

With the emergence of massively parallel processing units, parallelization has become a desirable property for new sequence models. The ability to parallelize the processing of sequences with respect to the sequence leng…

On the Importance of Multistability for Horizon Generalization in Reinforcement Learning

2026-05-12 · Asad Bakija, Florent De Geeter, Julien Brandoit, Pierre Sacré 외 arxiv

In reinforcement learning (RL), agents acting in partially observable Markov decision processes (POMDPs) must rely on memory, typically encoded in a recurrent neural network (RNN), to integrate information from past obse…

Reinforcement Learning

Were RNNs All We Needed?

2024-10-02 · Leo Feng, Frederick Tung, Mohamed Osama Ahmed, Yoshua Bengio 외

The introduction of Transformers in 2017 reshaped the landscape of deep learning. Originally proposed for sequence modelling, Transformers have since achieved widespread success across various domains. However, the scala…

AllMamba

Resurrecting Recurrent Neural Networks for Long Sequences

2023-03-11 · Antonio Orvieto, Samuel L Smith, Albert Gu, Anushan Fernando 외

Recurrent Neural Networks (RNNs) offer fast inference on long sequences but are hard to optimize and slow to train. Deep state-space models (SSMs) have recently been shown to perform remarkably well on long sequence mode…

ClassificationComputational EfficiencySequential Image ClassificationState Space Models

Fixed-Point RNNs: From Diagonal to Dense in a Few Iterations

2025-03-13 · Sajad Movahedi, Felix Sarnthein, Nicola Muca Cirone, Antonio Orvieto

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,…

MambaState Space Models