Real Time Recurrent Learning

Real-time recurrent learning (RTRL) is an online training algorithm for recurrent neural networks (RNNs) aiming to overcome the limitations of backpropagation through time (BPTT) by enabling continuous learning from sequential data without needing to store past activations. Current research focuses on improving RTRL's computational efficiency through architectural modifications like linear recurrent units and sparse networks, as well as exploring its application in reinforcement learning and various signal processing tasks, including speech and biomedical signal analysis. The development of efficient RTRL algorithms holds significant promise for biologically plausible neural network models and real-time applications requiring continuous adaptation to dynamic environments.

Papers

September 17, 2024

Bio-Inspired Mamba: Temporal Locality and Bioplausible Learning in Selective State Space Models
Jiahao Qin
Neuromorphic Hardware Biological Neural Network Spike Timing Dependent Plasticity Selective State Space Plausible Learning Multi Modal Mamba Twister Block Biological Learning Real Time Recurrent Learning

September 2, 2024

Real-Time Recurrent Learning using Trace Units in Reinforcement Learning
Esraa Elelimy, Adam White, Michael Bowling, Martha White
Reinforcement Learning Recurrent Neural Network Fuzzy Trace Recurrent Unit Linear Recurrent Real Time Recurrent Learning

February 22, 2024

Quaternion recurrent neural network with real-time recurrent learning and maximum correntropy criterion
Pauline Bourigault, Dongpo Xu, Danilo P. Mandic
Neural Network Quaternion Valued Convolutional Neural Network Real Time Recurrent Learning Maximum Correntropy Criterion Quaternion Property Correntropy Learning

November 8, 2023

Real-Time Recurrent Reinforcement Learning
Julian Lemmel, Radu Grosu
Actor Critic Algorithm Real Time Recurrent Learning

November 7, 2023

Estimating Post-Synaptic Effects for Online Training of Feed-Forward SNNs
Thomas Summe, Clemens JS Schaefer, Siddharth Joshi
Neural Network Back Propagation Spatio Temporal Learning Online Training Real Time Recurrent Learning

May 30, 2023

Exploring the Promise and Limits of Real-Time Recurrent Learning
Kazuki Irie, Anand Gopalakrishnan, Jürgen Schmidhuber
Recurrent Neural Network Continuum Limit Recurrent Network Actor Critic Promise Real Time Recurrent Learning

March 10, 2023

Efficient Real Time Recurrent Learning through combined activity and parameter sparsity
Anand Subramoney
Recurrent Neural Network Back Propagation Recurrent Network Generating User Activity Parameter Efficient Sparsity Real Time Recurrent Learning

January 20, 2023

Scalable Real-Time Recurrent Learning Using Columnar-Constructive Networks
Khurram Javed, Haseeb Shah, Rich Sutton, Martha White
Recurrent Neural Network Back Propagation Recurrent Network Real Time Recurrent Learning Geometric Neural Network

November 8, 2022

Improving performance of real-time full-band blind packet-loss concealment with predictive network
Viet-Anh Nguyen, Anh H. T. Nguyen, Andy W. H. Khong
System Performance Processing Pipeline Speech Distortion Packet Loss Concealment Real Time Recurrent Learning Prediction Network

July 13, 2022

Prediction of the motion of chest internal points using a recurrent neural network trained with real-time recurrent learning for latency compensation in lung cancer radiotherapy
Michel Pohl, Mitsuru Uesaka, Kazuyuki Demachi, Ritu Bhusal Chhatkuli
Recurrent Neural Network Optical Flow Exceptional Point Deformable Image Registration Radiation Oncology Respiratory Motion Real Time Recurrent Learning 3D Tumor

January 27, 2022

An Empirical Analysis of Recurrent Learning Algorithms In Neural Lossy Image Compression Systems
Ankur Mali, Alexander Ororbia, Daniel Kifer, Lee Giles
Recurrent Neural Network Image Compression Empirical Analysis Neural Compression Unbiased Gradient Real Time Recurrent Learning

Real Time Recurrent Learning

Papers

Bio-Inspired Mamba: Temporal Locality and Bioplausible Learning in Selective State Space Models

Real-Time Recurrent Learning using Trace Units in Reinforcement Learning

Quaternion recurrent neural network with real-time recurrent learning and maximum correntropy criterion

Real-Time Recurrent Reinforcement Learning

Estimating Post-Synaptic Effects for Online Training of Feed-Forward SNNs

Exploring the Promise and Limits of Real-Time Recurrent Learning

Efficient Real Time Recurrent Learning through combined activity and parameter sparsity

Scalable Real-Time Recurrent Learning Using Columnar-Constructive Networks

Improving performance of real-time full-band blind packet-loss concealment with predictive network

Prediction of the motion of chest internal points using a recurrent neural network trained with real-time recurrent learning for latency compensation in lung cancer radiotherapy

An Empirical Analysis of Recurrent Learning Algorithms In Neural Lossy Image Compression Systems