23/04/2026
A 28×10 rectangular lattice of tanh-filter cells, trained online with plain SGD at batch size 1, no learned encoder or decoder, 52% on row-wise MNIST.
18/12/2025
Making IIR filter coefficients input-dependent without breaking the elegance of real-time recurrent learning.
5/10/2025
How does equilibrium propagation work in the complex domain?
14/12/2024
The physics of DIMPLE and some unexpected properties