Adaptive Neural Dynamics
DREAM: Dynamic Recall and Elastic Adaptive Memory
A PyTorch implementation of continuous-time RNN cells with surprise-driven plasticity and liquid time-constants for adaptive neural dynamics.
Key Features
Surprise-Driven Plasticity
The cell adapts its learning rate based on prediction error surprise
Liquid Time-Constants
Integration speeds change dynamically based on input novelty
Fast Weights
Low-rank weight decomposition enables efficient meta-learning
Sleep Consolidation
Memory stabilization during low-surprise periods
Use Cases
Online Learning
Continuously adapting to new patterns in real-time
Non-Stationary Sequences
Handle data distributions that change over time
Speech Recognition
Model acoustic patterns with temporal dynamics
Few-Shot Learning
Rapid adaptation to new tasks with minimal data
Time Series Forecasting
Predict future values with adaptive dynamics
Anomaly Detection
Detect unusual patterns using reconstruction error
Ready to Get Started?
Explore the documentation and start building with DREAM today.
View Documentation