Computational modeling methodology for chronic stress as excitatory-inhibitory (E/I) balance perturbation in recurrent working-memory networks. Use when modeling stress effects on prefrontal cortex, studying resilience mechanisms, or analyzing E/I balance disruptions in neuropsychiatric conditions.
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Um comando direto ignora o prompt de revisão. Verifique a origem antes de executá-lo.
Instruções da origem · Visualização somente leitura
name
chronic-stress-ei-balance-rnn
description
Computational modeling methodology for chronic stress as excitatory-inhibitory (E/I) balance perturbation in recurrent working-memory networks. Use when modeling stress effects on prefrontal cortex, studying resilience mechanisms, or analyzing E/I balance disruptions in neuropsychiatric conditions.
Key Finding:
Only S↑[W_{I→E}] (stronger I→E synapses) reproduces all three signatures simultaneously, suggesting a causal cascade where enhanced inhibition leads to excitatory hypofunction as downstream consequence.
3. Resilience Analysis
Training Protocol:
Naive networks: trained without stress operator
Resilient networks: trained with stress operator applied
Compare performance, connectivity, dynamics, energy
Resilience Trade-offs:
✓ Preserves task performance under stress
✓ Confines network to same dynamical subspace with/without stress
✓ Maintains energetic regime
✗ Reduced generalization to longer delays (out-of-distribution)
✗ Decreased network density and reciprocity
✗ Shift toward directional (non-symmetric) information flow
Interpretation:
Resilience = specialized solution tuned to training regime → computational analogue of behavioral rigidity/habit formation observed in chronic stress animal models.
4. Analysis Metrics
E/I Balance Quantification:
# Excitatory drive (recurrent input to E population)
h_E(t) = W_{EE} r_E(t) + W_{EI} r_I(t)
# Inhibitory drive (recurrent input to I population)
h_I(t) = W_{IE} r_E(t) + W_{II} r_I(t)
# E/I balance ratio
EI_ratio = mean(h_E) / mean(h_I)
# Inhibitory dominance: EI_ratio < 1
Network Topology:
Density: fraction of non-zero connections
Reciprocity: symmetry of bidirectional connections (W_{ij} vs W_{ji})
Effective connectivity: source-specific decomposition of synaptic cost
Dynamical Analysis:
Geometric dynamics: manifold structure of population activity
Energy landscape: E(r) = -1/2 r^T W r + I^T r
Subspace preservation: principal component overlap between conditions
This implies different experimental observations (elevated PV activity, more GABAergic contacts, reduced excitation) may form a causal cascade rather than parallel mechanisms.
2. Resilience-Generalization Trade-off
Resilient networks:
Preserve function under stress (robust)
But lose flexibility outside training distribution (rigid)
Neurobiological parallel:
Chronic stress patients show preserved routine function but impaired adaptation to novel situations → computational analogue of behavioral rigidity.
3. Energetic Signature
Resilient networks maintain same energy landscape geometry with/without stress, suggesting:
