synaptic-intelligence

Continual learning baselines and strategies from popular papers, using Avalanche. We include EWC, SI, GEM, AGEM, LwF, iCarl, GDumb, and other strategies.

A brain-inspired version of generative replay for continual learning with deep neural networks (e.g., class-incremental learning on CIFAR-100; PyTorch code).

Continual Hyperparameter Selection Framework. Compares 11 state-of-the-art Lifelong Learning methods and 4 baselines. Official Codebase of "A continual learning survey: Defying forgetting in classification tasks." in IEEE TPAMI.

PyTorch implementation of a VAE-based generative classifier, as well as other class-incremental learning methods that do not store data (DGR, BI-R, EWC, SI, CWR, CWR+, AR1, the "labels trick", SLDA).

GenPark AI Agent Skill - Knowledge distillation teacher-student loss engine with temperature-scaled Kullback-Leibler divergence for continual model compression and capability preservation.

GenPark AI Agent Skill - Continual learning experience replay buffer with streaming reservoir sampling and priority weighting to balance past vs new agent trajectories.

GenPark AI Agent Skill - Gradient Episodic Memory (GEM / A-GEM) projection engine constraining parameter updates to prevent interference with past task gradients.

GenPark AI Agent Skill - Elastic Weight Consolidation (EWC) engine calculating Fisher Information Matrix diagonal penalties to prevent catastrophic forgetting during sequential task learning.

GenPark AI Agent Skill - Elastic Weight Consolidation (EWC) engine calculating Fisher Information Matrix diagonal penalties to prevent catastrophic forgetting during sequential task learning.

GenPark AI Agent Skill - Gradient Episodic Memory (GEM / A-GEM) projection engine constraining parameter updates to prevent interference with past task gradients.

GenPark AI Agent Skill - Continual learning experience replay buffer with streaming reservoir sampling and priority weighting to balance past vs new agent trajectories.

GenPark AI Agent Skill - Synaptic Intelligence (SI) online path integral engine accumulating parameter trajectory contributions to protect vital synaptic pathways across tasks.

GenPark AI Agent Skill - Synaptic Intelligence (SI) online path integral engine accumulating parameter trajectory contributions to protect vital synaptic pathways across tasks.

Implements regularization-based continual learning strategies that mitigate catastrophic forgetting by penalizing large parameter changes. Includes reproducible implementations of EWC, Synaptic Intelligence (SI), and Memory Aware Synapses (MAS) with experiment scripts and benchmark evaluations on Split-MNIST, Permuted-MNIST, and CIFAR-100.

This project investigates various continual learning methods to mitigate catastrophic forgetting

🚀 Clear your terminal instantly with cls, a fast C utility that removes all previous output and scrollback history on macOS and Linux.

A comparative evaluation of continual learning strategies to mitigate catastrophic forgetting in neural networks.