# Learning in Feedback-driven Recurrent Spiking Neural Networks using full-FORCE Training

arXiv: https://arxiv.org/abs/2205.13585 · HTML: https://mlsanity.com/paper/2205.13585/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 24 of its 63 references that are in the map.
- First posted on arXiv 26 May 2022.

## Builds on (most cited first, 24 of 24)

- [Next Generation Reservoir Computing](https://mlsanity.com/paper/2106.07688/) (2021)
- [Temporal Spike Sequence Learning via Backpropagation for Deep Spiking Neural Networks](https://mlsanity.com/paper/2002.10085/) (2020)
- [Gated Feedback Recurrent Neural Networks](https://mlsanity.com/paper/1502.02367/) (2015)
- [A scalable multi-core architecture with heterogeneous memory structures for Dynamic Neuromorphic Asynchronous Processors (DYNAPs)](https://mlsanity.com/paper/1708.04198/) (2017)
- [T2FSNN: Deep Spiking Neural Networks with Time-to-first-spike Coding](https://mlsanity.com/paper/2003.11741/) (2020)
- [Noisy Recurrent Neural Networks](https://mlsanity.com/paper/2102.04877/) (2021)
- [Lyapunov spectra of chaotic recurrent neural networks](https://mlsanity.com/paper/2006.02427/) (2020)
- [Spike-Train Level Backpropagation for Training Deep Recurrent Spiking Neural Networks](https://mlsanity.com/paper/1908.06378/) (2019)
- [full-FORCE: A Target-Based Method for Training Recurrent Networks](https://mlsanity.com/paper/1710.03070/) (2017)
- [Implementing Spiking Neural Networks on Neuromorphic Architectures: A Review](https://mlsanity.com/paper/2202.08897/) (2022)
- [Supervised Learning in Spiking Neural Networks with FORCE Training](https://mlsanity.com/paper/1609.02545/) (2016)
- [Mapping Spiking Neural Networks to Neuromorphic Hardware](https://mlsanity.com/paper/1909.01843/) (2019)
- [Unsupervised Heart-rate Estimation in Wearables With Liquid States and A Probabilistic Readout](https://mlsanity.com/paper/1708.05356/) (2017)
- [Mapping of Local and Global Synapses on Spiking Neuromorphic Hardware](https://mlsanity.com/paper/1908.08024/) (2019)
- [PyCARL: A PyNN Interface for Hardware-Software Co-Simulation of Spiking Neural Network](https://mlsanity.com/paper/2003.09696/) (2020)
- [Compiling Spiking Neural Networks to Neuromorphic Hardware](https://mlsanity.com/paper/2004.03717/) (2020)
- [NeuroXplorer 1.0: An Extensible Framework for Architectural Exploration with Spiking Neural Networks](https://mlsanity.com/paper/2105.01795/) (2021)
- [A Design Flow for Mapping Spiking Neural Networks to Many-Core Neuromorphic Hardware](https://mlsanity.com/paper/2108.12444/) (2021)
- [Enabling Resource-Aware Mapping of Spiking Neural Networks via Spatial Decomposition](https://mlsanity.com/paper/2009.09298/) (2020)
- [On the Role of System Software in Energy Management of Neuromorphic Computing](https://mlsanity.com/paper/2103.12231/) (2021)
- [DFSynthesizer: Dataflow-based Synthesis of Spiking Neural Networks to Neuromorphic Hardware](https://mlsanity.com/paper/2108.02023/) (2021)
- [Design of Many-Core Big Little μ Brain for Energy-Efficient Embedded Neuromorphic Computing](https://mlsanity.com/paper/2111.11838/) (2021)
- [Design Technology Co-Optimization for Neuromorphic Computing](https://mlsanity.com/paper/2110.08131/) (2021)
- [Design-Technology Co-Optimization for NVM-based Neuromorphic Processing Elements](https://mlsanity.com/paper/2203.05311/) (2022)

## Built on by (most cited first, 2 of 2)

- [Heterogeneous Neuronal and Synaptic Dynamics for Spike-Efficient Unsupervised Learning: Theory and Design Principles](https://mlsanity.com/paper/2302.11618/) (2023)
- [Exploiting Heterogeneity in Timescales for Sparse Recurrent Spiking Neural Networks for Energy-Efficient Edge Computing](https://mlsanity.com/paper/2407.06452/) (2024)

## Abstract (arXiv)

Feedback-driven recurrent spiking neural networks (RSNNs) are powerful computational models that can mimic dynamical systems. However, the presence of a feedback loop from the readout to the recurrent layer de-stabilizes the learning mechanism and prevents it from converging. Here, we propose a supervised training procedure for RSNNs, where a second network is introduced only during the training, to provide hint for the target dynamics. The proposed training procedure consists of generating targets for both recurrent and readout layers (i.e., for a full RSNN system). It uses the recursive least square-based First-Order and Reduced Control Error (FORCE) algorithm to fit the activity of each layer to its target. The proposed full-FORCE training procedure reduces the amount of modifications needed to keep the error between the output and target close to zero. These modifications control the feedback loop, which causes the training to converge. We demonstrate the improved performance and noise robustness of the proposed full-FORCE training procedure to model 8 dynamical systems using RSNNs with leaky integrate and fire (LIF) neurons and rate coding. For energy-efficient hardware implementation, an alternative time-to-first-spike (TTFS) coding is implemented for the full- FORCE training procedure. Compared to rate coding, full-FORCE with TTFS coding generates fewer spikes and facilitates faster convergence to the target dynamics.