The Fragile Memory of Neural Networks, and the Metrics We Trust
19 Mar 2026
New research shows optimizer choice and metrics significantly impact catastrophic forgetting in neural networks—challenging common AI training practices.
Why Adam May Be Hurting Your Neural Network’s Memory
19 Mar 2026
Optimizer choice significantly impacts catastrophic forgetting in neural networks, with SGD outperforming Adam in retention and relearning tasks.
Teaching Machines to Remember Means Choosing What They Forget
18 Mar 2026
New research shows how optimizers like SGD, RMSProp, and Adam shape AI memory, revealing key tradeoffs between retention and relearning.
This Is How Your Model Forgets What It Just Learned
18 Mar 2026
Experiments across MNIST and RL tasks reveal how neural networks forget—and how optimizers like Adam and SGD impact memory retention.
Measuring Catastrophic Forgetting in AI
18 Mar 2026
How do neural networks forget? Explore retention, relearning, and activation overlap—the key metrics for measuring catastrophic forgetting in AI.
Study Finds Optimizer Choice Significantly Impacts Model Retention
18 Mar 2026
New research shows optimizer choice significantly impacts catastrophic forgetting in machine learning, challenging long-held assumptions about model retention.
Does the Adam Optimizer Amplify Catastrophic Forgetting?
17 Mar 2026
New research shows optimizer choice and flawed metrics may be distorting how we measure catastrophic forgetting in neural networks.