Backward Learning

Backward learning, a growing area of machine learning research, focuses on developing algorithms that learn from data by propagating information in the reverse direction of a typical forward pass, often circumventing the limitations of backpropagation. Current research explores various approaches, including pseudoinverse methods, models that leverage forward and backward trajectories to learn policies or dynamical systems, and architectures incorporating inverse evolution layers for regularization or constraint recovery in inverse problems. These advancements offer potential improvements in training efficiency, model interpretability, and the ability to handle complex or non-standard data structures, impacting fields ranging from reinforcement learning to inverse modeling and neural network design.

Papers

June 27, 2024

Semi-adaptive Synergetic Two-way Pseudoinverse Learning System
Binghong Liu, Ziqi Zhao, Shupan Li, Ke Wang
Neural Network Backward Learning Non Convex Gradient Descent

December 8, 2023

Backward Learning for Goal-Conditioned Policies
Marc Höftmann, Jan Robine, Stefan Harmeling
Reinforcement Learning Imitation Learning Goal Conditioned Policy Maze Environment Neural Network Policy Backward Learning

October 24, 2023

Minimax Forward and Backward Learning of Evolving Tasks with Performance Guarantees
Verónica Álvarez, Santiago Mazuelas, Jose A. Lozano
Incremental Learning Classification Task Minimax Optimality Performance Guarantee Backward Learning Drift Adaptation Minimax Risk Classifier

July 14, 2023

Inverse Evolution Layers: Physics-informed Regularizers for Deep Neural Networks
Chaoyu Liu, Zhonghua Qiao, Chao Li, Carola-Bibiane Schönlieb
Deep Neural Network Regularization Model Evolutionary Model Backward Learning Physic Based Regularization Heat Diffusion

May 14, 2023

Inverse Reinforcement Learning With Constraint Recovery
Nirjhar Das, Arpan Chattopadhyay
Inverse Reinforcement Learning Constrained Markov Decision Process Non Convex Optimization Problem Backward Learning Constraint Acquisition

April 15, 2023

Critical Sampling for Robust Evolution Operator Learning of Unknown Dynamical Systems
Ce Zhang, Kailiang Wu, Zhihai He
Dynamical System Evolutionary Operator Backward Learning

January 19, 2023

On backpropagating Hessians through ODEs
Axel Ciceri, Thomas Fischbacher
Neural ODE Hessian Matrix Second Order Ordinary Differential Equation Backward Learning

November 25, 2022

Mixture Manifold Networks: A Computationally Efficient Baseline for Inverse Modeling
Gregory P. Spell, Simiao Ren, Leslie M. Collins, Jordan M. Malof
Mixture Density Network Inverse Modeling Efficient Baseline Backward Learning INverse Problem

October 28, 2022

Meta-Learning Biologically Plausible Plasticity Rules with Random Feedback Pathways
Navid Shervani-Tabar, Robert Rosenbaum
Back Propagation Plasticity Rule Feedback Network Backward Learning

July 27, 2022

Investigating the Impact of Backward Strategy Learning in a Logic Tutor: Aiding Subgoal Learning towards Improved Problem Solving
Preya Shabrina, Behrooz Mostafavi, Mark Abdelshiheed, Min Chi, Tiffany Barnes
Global Impact Problem Solving Subgoal Search Backward Learning Problem Solving Strategy

April 4, 2022

Signal Propagation: A Framework for Learning and Inference In a Forward Pass
Adam Kohan, Edward A. Rietman, Hava T. Siegelmann
LeArning Abstract New Framework Scientific Inference Back Propagation Propagation Environment Forward Pas Backward Learning Learning Signal Hardware Efficient

March 20, 2022

Document Dewarping with Control Points
Guo-Wang Xie, Fei Yin, Xu-Yao Zhang, Cheng-Lin Liu
Document Image Backward Learning Document Dewarping

January 31, 2022

Learning on Arbitrary Graph Topologies via Predictive Coding
Tommaso Salvatori, Luca Pinchetti, Beren Millidge, Yuhang Song, Tianyi Bao, Rafal Bogacz, Thomas Lukasiewicz
LeArning Abstract Back Propagation Predictive Coding Arbitrary Topology Brain Like Backward Learning

November 26, 2021

Cyclic Graph Attentive Match Encoder (CGAME): A Novel Neural Network For OD Estimation
Guanzhou Li, Yujing He, Jianping Wu, Duowei Li
Neural Network Graph Attention Origin Destination Backward Learning OD Matrix

Backward Learning

Papers

Semi-adaptive Synergetic Two-way Pseudoinverse Learning System

Backward Learning for Goal-Conditioned Policies

Minimax Forward and Backward Learning of Evolving Tasks with Performance Guarantees

Inverse Evolution Layers: Physics-informed Regularizers for Deep Neural Networks

Inverse Reinforcement Learning With Constraint Recovery

Critical Sampling for Robust Evolution Operator Learning of Unknown Dynamical Systems

On backpropagating Hessians through ODEs

Mixture Manifold Networks: A Computationally Efficient Baseline for Inverse Modeling

Meta-Learning Biologically Plausible Plasticity Rules with Random Feedback Pathways

Investigating the Impact of Backward Strategy Learning in a Logic Tutor: Aiding Subgoal Learning towards Improved Problem Solving

Signal Propagation: A Framework for Learning and Inference In a Forward Pass

Document Dewarping with Control Points

Learning on Arbitrary Graph Topologies via Predictive Coding

Cyclic Graph Attentive Match Encoder (CGAME): A Novel Neural Network For OD Estimation