MesoNet: spectral diagnostics and pruning
eigenvalues / pruning / runaway updates. weight matrix isn't automatically the full Jacobian, especially w/ thresholds + resets in there
recorded exploration: eigenvalues of the weight matrix, dynamic pruning of weak connections, and alternative inverse-distance/exponential weighting schemes. The notebook associates large spectral values with runaway plasticity, but gives no controlled table or complete stability derivation in this entry.
important distinction: the weight matrix is not automatically the Jacobian of the full spiking/adaptive system. Thresholds, resets, inhibitory pathways and plasticity state can all change the dynamics being analyzed.
Eigenvalue distribution
Bifurcation and Lyapunov notes
References saved in this journal section
- review questions: which matrix is analyzed, at what operating point and time? Compare spectra before/after pruning with firing-rate and weight histories. Check whether a pruning threshold merely silences the network. Preserve the exact configuration before treating a spectral observation as a general stability criterion.