| name | mean-field-oscillatory-low-rank-rnn |
| description | Mean-field theory for rich oscillatory dynamics in low-rank recurrent networks with activity-dependent adaptation. Analyzes how low-rank structure and adaptation interact to produce complex oscillatory and chaotic behavior in recurrent neural networks. |
| trigger_words | ["mean field oscillatory","low rank recurrent","activity dependent adaptation","oscillatory dynamics RNN","rich oscillatory","RNN mean field theory"] |
| category | neuroscience |
Overview
This paper (arXiv:2606.30366) develops mean-field theory for recurrent neural networks with both low-rank structure and activity-dependent adaptation. Shows how these two mechanisms interact to produce rich dynamical regimes including oscillations, chaos, and multistability.
Core Theory
Model Structure
dx/dt = -x + W_lowrank * φ(x) + W_full * φ(x) - g_adapt * a * φ(x)
da/dt = (-a + φ(x)) / τ_adapt
- W_lowrank: Low-rank structured connectivity (rank r << N)
- W_full: Random full-rank connectivity (chaotic driver)
- g_adapt: Adaptation strength controlling oscillation onset
- τ_adapt: Adaptation timescale
Mean-Field Reduction
- Project dynamics onto low-rank subspace
- Use dynamical mean-field theory for random component
- Derive closed equations for order parameters
- Analyze fixed points, limit cycles, and chaotic regimes
Key Regimes
- Fixed point regime: Low adaptation, dominated by low-rank structure
- Oscillatory regime: Moderate adaptation creates limit cycles
- Chaotic regime: Strong random connectivity drives chaos
- Mixed regime: Adaptation + chaos produces complex metastable dynamics
Implementation
Mean-Field Equations
m_dot = f(m, q, a) # Order parameter dynamics
q_dot = g(m, q, a) # Variance dynamics
a_dot = (-a + h(m)) / τ_adapt # Adaptation dynamics
Bifurcation Analysis
- Vary g_adapt and σ_W (random connectivity strength)
- Track fixed point stability via Jacobian eigenvalues
- Identify Hopf bifurcations leading to oscillations
- Map chaos onset via largest Lyapunov exponent
Pitfalls
- Low-rank assumption: Results depend on r << N; for higher rank, mean-field breaks down
- Timescale separation: Analysis assumes τ_adapt >> 1; fast adaptation needs different treatment
- Activation function: Results derived for specific φ(x); different nonlinearities change dynamics
Applications
- Working memory: Oscillatory regimes support persistent activity
- Decision making: Multistability enables choice between alternatives
- Motor control: Oscillations drive rhythmic motor patterns
Verification
- Simulate full RNN and compare with mean-field predictions
- Verify bifurcation diagram matches numerical continuation
- Test different activation functions (ReLU, tanh, sigmoid)
- Validate low-rank assumption by varying rank r
Activation
mean field theory, oscillatory dynamics, low rank RNN, activity dependent adaptation, bifurcation analysis, RNN dynamics, working memory, neural oscillations