Publications, preprints and talks

Exploring the Geometry and Topology of Neural Network Loss Landscapes

Stefan Horoi*, Jessie Huang*, Bastian Rieck, Guillaume Lajoie, Guy Wolf^†, Smita Krishnaswamy^†

Proceedings of the 20th Symposium on Intelligent Data Analysis (IDA), Springer’s LNCS vol. 13205, 2022

We presented a novel way of sampling the loss landscape of artificial neural networks and use manifold learning and computational homology to study the sampled data. We found that the extracted geometrical and topological features of the loss landscapes hold meaningful information about the model’s ability to generalize.

On the Inadequacy of CKA as a Measure of Similarity in Deep Learning

MohammadReza Davari*, Stefan Horoi*, Amine Natik, Guillaume Lajoie, Guy Wolf^†, Eugene Belilovsky^†

Poster presentation, GTRL workshop at ICLR, 2022

Goal-driven optimization of single-neuron properties in artificial networks reveals regularization role of neural diversity and adaptation

Victor Geadah, Stefan Horoi, Giancarlo Kerg, Guillaume Lajoie, Guy Wolf

Preprint, 2022

Top-down optimization recovers biological coding principles of single-neuron adaptation in RNNs

Victor Geadah, Stefan Horoi, Giancarlo Kerg, Guillaume Lajoie, Guy Wolf

Poster presentation, CoSyNe, 2022

Visualizing High-Dimensional Trajectories on the Loss-Landscape of ANNs

Stefan Horoi*, Jessie Huang*, Guy Wolf^†, Smita Krishnaswamy^†

Poster presentation, DLIG workshop at NeurIPS, 2020

Low-dimensional dynamics of encoding and learning in recurrent neural networks

Stefan Horoi, Victor Geadah, Guillaume Lajoie, Guy Wolf

Proceedings of the 33rd Canadian Conference on Artificial Intelligence (CAIAC), Springer’s LNCS vol. 12109, 2020