Anamika Agrawal

Using simple DNNs with dendrite-inspired architectures, we show that single neurons exhibit a morphology-driven phase transition in learnability, with robustness–adaptability trade-offs that suggest new inductive biases for NeuroAI model design.