Multi-Frequency Local Plasticity for Visual Representation Learning

ArXi:2604.09734v1 Announce Type: cross We study how far structured architectural bias can compensate for the absence of end-to-end gradient-based representation learning in visual recognition. Building on the VisNet tradition, we Representational layers are trained without end-to-end backpropagation through the full hierarchy; only the final linear readout and top-down projection matrices are optimized by gradient descent. We. therefore. interpret the model as a hybrid system that is predominantly locally trained but includes a small number of gradient-trained parameters.