Topological-Reinforcement-Operator

The Topological Reinforcement Operator (ORT) is an innovative computational framework designed to emulate memory consolidation processes in complex neural networks. This project bridges theoretical concepts and practical applications, ranging from simplified models to detailed explorations of human connectomics.

Overview

The ORT reveals how functional and resilient memory can emerge directly from the topological organization of a neural network. By demonstrating a principle of computational parsimony with biological plausibility, this framework opens new avenues for understanding memory dynamics in biological systems.

Visualization of a sample engram from the human connectome, illustrating a "core-periphery" structure and a "rich-club" organization.

Key Findings

• Dual Principle (Critical Mass vs. Elite): Biological networks optimize memory using "elite" cores (P95), while information networks require a "critical mass" (P90) for resilience.
• Topological Specificity: Memory function collapses (F1-score ≈ 0) when the precise configuration of the brain network is altered, indicating that not only the hubs matter, but also how they are interconnected.
• Computational Parsimony: The degree-based ORT operates ~96 times faster and uses ~19 times less RAM than traditional methods like PageRank, while achieving comparable functional performance.
• Biological Plausibility: The ORT achieves an outstanding performance (F1-score ~87%) in memory recovery tasks using the human connectome.

Interactive Experiments

Data Requirements

To conduct experiments with biological connectomes:

1. Download the .edges file from the original source.
2. Upload it to Google Colab before execution.

• Macaque Connectome: Network Repository
• Human Connectome: Network Repository

Experimental Pipeline

1. Fundamental Experiment: ORT in Networks and Connectomes
   A complete pipeline from GNN to validation across multiple datasets.
   Open In Colab
2. Journey into the Brain: ORT in the Human Connectome
   Direct application to a digitized human brain.
   Open In Colab

Repository Structure

• /Notebooks: All experiment notebooks (English and Spanish)
• /Reports: Complete research articles (English and Spanish)
• /outputs: Result data generated from experiments
  • /csv: Quantitative data in CSV format
  • /img: Visualizations and graphs

This project exemplifies independent and open science, underscoring that cutting-edge research can emerge from accessible and non-institutional environments.

Support and Visibility

Supporting this research enhances its visibility within the scientific community. Simple actions like starring the repository on GitHub, sharing on social media, or citing the work can significantly impact the reach of this project.