With James Maguire, I explored a neuron architecture that learns transmission delays rather than treating time as a fixed sequence of inputs. The manuscript asks how repeated patterns can be detected and segmented within a noisy signal without supervised labels.
The question that remains
The reported experiments used two simple patterns. They are an early demonstration, not evidence of general language understanding. The manuscript, NYAS poster, figures and code are public; the manuscript is not an accepted NIPS paper.


