MATH · IN · MODELS

Murray 2023 Continuous-Time RNN for Pattern Recognition

Research (Murray 2023; custom-trained from scratch on a modular-arithmetic / SET-inspired pattern task)

Structures found in this family (1)

Limit cycle (stable periodic attractor)

By model (1)

Murray 2023 continuous-time RNN (100 recurrent units) · 100 units
Limit cycle (stable periodic attractor)

Observations (1)

A continuous-time RNN trained on a modulo-3 arithmetic task can learn a low-dimensional limit-cycle phase code whose input-induced phase shifts implement the computation (one solution among several)

Papers

Phase codes emerge in recurrent neural networks optimized for modular arithmetic (2023)