Pehlevan Group

Our research areas are theoretical neuroscience, machine learning, and theory of neural computation in natural and artificial systems. We are widely interested in how computation arises in neural networks from the collective dynamics of simple processing units, or neurons.

Latest Publications and Preprints

B. Bordelon, A. Atanasov, and C. Pehlevan, “A Dynamical Model of Neural Scaling Laws ,” arXiv:2402.01092, 2024.

T. Kumar, B. Bordelon, S. J. Gershman, and C. Pehlevan, “Grokking as the Transition from Lazy to Rich Training Dynamics,” International Conference on Learning Representations (ICLR), 2024.

PDF

B. Bordelon*, L. Noci*, M. B. Li, B. Hanin*, and C. Pehlevan*, “Depthwise Hyperparameter Transfer in Residual Networks: Dynamics and Scaling Limit,” International Conference on Learning Representations (ICLR), 2024.

PDF

B. Bordelon*, L. Noci*, M. Li, B. Hanin*, and C. Pehlevan*, “Depthwise Hyperparameter Transfer in Residual Networks: Dynamics and Scaling Limit,” in NeurIPS 2023 Workshop on Mathematics of Modern Machine Learning, 2023.

PDF

H. T. Chaudhry, J. A. Zavatone-Veth, D. Krotov, and C. Pehlevan, “Long Sequence Hopfield Memory,” in Associative Memory & Hopfield Networks in 2023 Workshop, NeurIPS, 2023.

PDF

N. M. Chapochnikov, C. Pehlevan, and D. B. Chklovskii, “Reply to Castro et al.: Do connectomes possess markers of activity-dependent synaptic plasticity?” Proceedings of the National Academy of Sciences (PNAS), vol. 120, no. 50, pp. e2317056120, 2023.

PDF

B. Bordelon and C. Pehlevan, “Self-Consistent Dynamical Field Theory of Kernel Evolution in Wide Neural Networks,” Journal of Statistical Mechanics: Theory and Experiment (Machine Learning Special Issue), pp. 114009, 2023.

PDF

B. Bordelon, P. Masset, H. Kuo, and C. Pehlevan, “Loss Dynamics of Temporal Difference Reinforcement Learning ,” in Advances in Neural Information Processing Systems (NeurIPS), 2023.

PDF