| name | holobrain-holograph-oscillatory-gnn |
| description | HoloBrain and HoloGraph framework: modeling brain rhythms through coupled oscillatory synchronization and applying this principle to graph neural networks. Addresses GNN over-smoothing and enables reasoning on graphs through oscillatory dynamics.
|
HoloBrain & HoloGraph: Oscillatory Synchronization for Brain Modeling and GNNs
Paper: arXiv:2602.00057
Authors: Tingting Dan, Jiaqi Ding, Guorong Wu
Categories: q-bio.NC, cs.LG
Year: 2026
Overview
Two-part framework connecting neuroscience and machine learning through oscillatory synchronization:
- HoloBrain: Models brain rhythms through interference of spontaneously synchronized neural oscillations
- HoloGraph: Applies synchronization principle to GNNs, enabling oscillatory computation beyond heat diffusion
Key Concepts
Neural Oscillatory Synchronization
- Brain rhythms emerge from synchronization of coupled neural oscillators
- Phase relationships between oscillators encode abstract concepts
- Synchronization patterns dynamically reconfigure for different cognitive functions
HoloGraph: Oscillatory GNNs
- Each node is an oscillator; edges define coupling strength
- Information propagation through phase synchronization rather than feature diffusion
- Addresses over-smoothing: oscillatory dynamics maintain distinct phase patterns even after many iterations
Methodology
HoloGraph Implementation
- Replace conventional GNN message passing with oscillatory synchronization
- Node states as complex numbers (amplitude + phase)
- Information encoded in phase relationships
- Synchronization dynamics enable iterative refinement
- Readout maps final phase patterns to predictions
Advantages over Traditional GNNs
- No over-smoothing
- Natural multi-scale representation
- Biological plausibility
- Enhanced reasoning capability
Applications
- Brain rhythm modeling
- Graph classification
- Molecular property prediction
- Knowledge graph reasoning
Key Insights
- Shared mechanism: same oscillatory synchronization for brain rhythms and graph computation
- Over-smoothing solution through oscillatory dynamics
- Phase as richer representation
- Biology inspires computation
References
- Dan, T., Ding, J., & Wu, G. (2026). HoloBrain & HoloGraph. arXiv:2602.00057.
