Memory-DD: A Low-Complexity Dendrite-Inspired Neuron for Temporal Prediction Tasks
Dongjian Yang, Xiaoyuan Li, Chuanmei Xi, Ye Sun, Gang Liu* · 2025
Read Memory-DD: A Low-Complexity Dendrite-Inspired Neuron for Temporal Prediction Tasks by Dongjian Yang, Xiaoyuan Li, Chuanmei Xi, Ye Sun, Gang Liu* (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-04094.txt.
Description
Neurons in brains grow branching dendrites. Artificial neurons modelled on them are cheap to compute and quick at inference. Yet they were built for still images, not sequences. Dongjian Yang and colleagues extend the idea to time series with Memory-DD, presented in a 2025 neural computation preprint. Their neuron aims at temporal prediction: sensor streams, financial series, and traffic flows.
Memory-DD couples two groups of dendrite-inspired neurons. Neither group uses a nonlinear activation function, yet together they realise nonlinear mappings and extract logical relationships between features across a sequence. On eighteen temporal classification benchmarks, the model beat the standard LSTM network on accuracy. On nine regression datasets it matched LSTM while using half the parameters and fewer operations. The authors present the design as carrying the dendritic family's low-cost advantages into sequence modelling, for both classification and regression jobs.
Catalog record
| CREATOR | Dongjian Yang, Xiaoyuan Li, Chuanmei Xi, Ye Sun, Gang Liu* |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 66 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-04094