Harvard University

Papers and research supported in part or in whole by The Swartz Foundation

N. Hiratani and H. Sompolinsky,
"Optimal quadratic binding for relational reasoning in vector symbolic neural architectures Neural Computation," (in press). arXiv: 2204.07186

M. Farrell, B. Bordelon, S. Trivedi, and C. Pehlevan,
"Capacity of Group-invariant Linear Readouts from Equivariant Representations: How Many Objects can be Linearly Classified Under All Possible Views?" International Conference on Learning Representations (2022).

M. Farrell, S. Recanatesi, G. Lajoie and E. Shea-Brown,
"Gradient-based learning drives robust representations in recurrent neural networks by balancing compression and expansion," Nature Machine Intelligence (2022).

N. Hiratani, Y. Mehta, T.P. Lillicrap, and P.E. Latham,
"On the stability and scalability of node perturbation learning," (presented at NeurIPS 2022)

Pakman, A., Nejatbakhsh, A., Gilboa, D., Makkej, A., Mazzucato, L., Wibral, M., Schneidman, E.,
"Estimating the Unique Information of Continuous Variables in Recurrent Networks", NeurIPS 2021.

Wang, T., Buchanan S., Gilboa, D., Wright, J.,
"Deep Networks Provably Classify Data on Curves", NeurIPS 2021.

Buchanan, S., Gilboa, D., Wright, J.,
"Deep Networks and the Multiple Manifold Problem", ICLR 2021.

Gilboa, D., Pakman, A., Vatter, T.,
"Marginalizable Density Models", Arxiv preprint, 2021.

J. Steinberg, M. Advani, and H. Sompolinsky.
"A new role for circuit expansion for learning in neural networks," Physical review. E, 01 Feb 2021, 103(2-1):022404.

Advani, Madhu S., Andrew M. Saxe, and Haim Sompolinsky.
"High-dimensional dynamics of generalization error in neural networks." Neural Networks 132 (2020): 428-446.

N. Shaham, J. Chandra, G. Kreiman, and H. Sompolinsky.
"Continual learning, replay and consolidation in a forgetful memory network model," Cosyne Abstracts 2020.

Julia Steinberg.
"Associative Memory of Structured Knowledge," APS Meeting, March, 2020.

Return to Swartz Program in Theoretical Neurobiology at Harvard main page

Return to main Research page