plasticity-network-framework

# Quantifying Plasticity: Network-Based Framework > **Paper**: "Quantifying plasticity: a network-based framework linking structure to dynamical regimes" (arXiv:2603.25180) > **Core Insight**: Plasticity operationalized as `system size / connectivity strength` ## Core Problem **Plasticity** is a fundamental property of complex systems (brain, organisms, ecosystems), but typically remains a **descriptive concept** inferred retrospectively from observed outcomes. This paper provides a **quantitative operational definition** linking network structure to dynamical regimes. ## Framework ### Plasticity Definition ``` Plasticity = System Size / Connectivity Strength Where: - System Size (N): Number of elements → determines state space dimensionality - Connectivity Strength (C): Coupling among elements → determines state coupling ``` ### Two Dynamical Regimes | Regime | Plasticity | System Size | Connectivity | Behavior | |--------|-----------|-------------|--------------|----------| | **High Plasticity** | High (N/C >> 1) | Large | Weak | Many accessible states, flexible responses | | **Low Plasticity (Rigid)** | Low (N/C << 1) | Small | Strong | Locked states, constrained dynamics | ### Mechanism 1. **System Size → Dimensionality** - More elements = larger state space (2^N possible configurations for binary elements) - High dimensionality enables diverse dynamical trajectories 2. **Connectivity → Coupling** - Stronger connections = tighter state coupling - High connectivity locks system into restricted state trajectories 3. **Ratio N/C → Plasticity** - Large N / weak C = flexible system (many reachable states) - Small N / strong C = rigid system (state locking) ## Applications ### Brain Plasticity - **Developmental plasticity**: Young brains (high N, weak C) → high flexibility - **Adult rigidity**: Mature brains (moderate N, strong C) → stable but less adaptable - **Pathology**: Stroke/injury → changes N/C ratio → altered plasticity ### Neural Networks - **Overparameterized models**: High N, moderate C → high plasticity (good for learning) - **Compact models**: Low N, strong C → rigidity (stable but limited adaptation) - **Training dynamics**: Plasticity affects optimization landscape ### Ecosystems - **Biodiversity (N)**: More species → larger state space - **Interaction strength (C)**: Strong trophic links → tighter coupling - **Resilience**: High plasticity ecosystems adapt to disturbances ## Reusable Patterns ### Pattern 1: Plasticity Assessment ``` Given complex system with N elements and connectivity matrix: 1. Compute connectivity strength C (average coupling strength) 2. Calculate plasticity ratio = N / C 3. Classify regime: - Ratio > threshold_high → plastic system - Ratio < threshold_low → rigid system - Intermediate → mixed behavior ``` ### Pattern 2: State Space Accessibility ``` For system with plasticity ratio: 1. High plasticity: Explore broad state space (good for exploration) 2. Low plasticity: Narrow state space (good for stability) 3. Trade-off: Balance flexibility vs robustness ``` ### Pattern 3: Intervention Design ``` To modify plasticity: - Increase system size N (add elements, increase dimensionality) - Decrease connectivity C (loosen coupling, increase independence) - Or both simultaneously ``` ## Quantification Methods ### Connectivity Strength Metrics 1. **Average coupling**: Mean interaction strength 2. **Network density**: Fraction of connected pairs 3. **Weighted connectivity**: Sum of edge weights / N 4. **Spectral measure**: Largest eigenvalue of adjacency matrix ### System Size Metrics 1. **Number of nodes**: Direct count of elements 2. **Effective dimensionality**: PCA on state vectors 3. **Configuration entropy**: log2(N) for binary elements ## Comparison with Existing Concepts | Concept | Focus | Quantification | Network Basis | |---------|-------|---------------|--------------| | Structural plasticity | Synapse formation/deletion | Yes (synapse count) | Partial | | Functional plasticity | Activity changes | Yes (signal metrics) | No | | **This framework** | **Structure-dynamics link** | **Yes (N/C ratio)** | **Full** | ## Pitfalls - **Size vs connectivity trade-off**: Increasing N often increases C → plasticity may not change monotonically - **Network topology ignored**: Ratio N/C assumes uniform connectivity → misses heterogeneous structure effects - **Dynamics oversimplified**: State space dimensionality depends on element dynamics, not just count - **Time-scale neglected**: Plasticity evolves over time → static ratio ignores temporal dynamics ## Related Skills - [[effective-plasticity]] - Network-based framework for plasticity quantification - [[structural-plasticity-growth-stability]] - Analysis of structural plasticity in neural networks - [[neural-manifold-dynamics-learning]] - Neural manifold learning dynamics - [[synaptic-weight-distributions-plasticity-geometry]] - Synaptic weight distributions and plasticity geometry ## Activation plasticity quantification, complex systems, network structure, dynamical regimes, brain resilience, neural rigidity, state space, connectivity strength, system dimensionality, ecosystem stability, N/C ratio