Measuring Catastrophic Forgetting in Neural Networks

arxiv:1708.02072

Original source · Paper JSON

Data, training, optimization · 2017-08-07

Cites (1) · Cited by (1)

Complete cleaned Markdown · Source Markdown including bibliography

Read in parts

1. Part 1
2. Part 2
3. Part 3
4. Part 4

Contents

• Measuring Catastrophic Forgetting in Neural Networks
• Abstract
• Introduction
• Problem Formulation
• Why Does Catastrophic Forgetting Occur?
• Previous Surveys
• Mitigating Catastrophic Forgetting
• Regularization Methods
• Ensemble Methods
• Rehearsal Methods
• Dual-Memory Models
• Sparse-Coding Methods
• Experimental Setup
• Dataset Description
• MNIST
• CUB-200
• AudioSet
• Models Evaluated
• Standard Multi-Layer Perceptron
• Elastic Weight Consolidation
• PathNet
• GeppNet
• Fixed Expansion Layer
• Experiments and Results
• Evaluation Metrics
• Experimental Results
• Data Permutation Experiment
• Incremental Class Learning
• Multi-Modal Experiment
• Discussion
• Conclusion
• Acknowledgements
• Appendix A Supplemental Material
• Training Parameters
• Reproduction Validation Experiments
• Plots and Tables for Experimental Results
• Data Permutation Experiment
• Incremental Class Learning Experiment
• Multi-Modal Experiment
• Fast Correlation Based Filter
• Ideal Model
• Instructions for reporting errors