Jun 2022: Debugging using Orthogonal Descent, Workshop on Updatable Machine Learning, ICML 2022

Oct 2021: Submitted work on scaling laws of LTI systems to ICLR; Started work on non-parametric convolution augmented transformer; (update: iclr rejected)

Aug 2021: MASc Completed.

May 2021: ICML accepted! Started work at ABR!

April 2021: Submitted thesis.

Feb 2021: Submitted work to ICML.

Nov 2020: ICLR rejected...rip.

Sept 2020: Submitted work to ICLR.

May 2020: Started using the (modified) LMU for NLP applications.

March 2020: Covid!!

Jan 2020: Started LMU work.

Sept 2019: Started MASc in Systems Design Engineering.

April 2018: Obtained bachelors in Mathematical Physics from the University of Waterloo.

Publications

Conference and Workshop Papers

• Narsimha Chilkuri, Chris Eliasmith (2022) Debugging using Orthogonal Gradient Descent. In Updatable Machine Learning, ICML. arXiv.  Abstract  PDF  arXiv
• Narsimha Reddy Chilkuri, Chris Eliasmith (2021) Parallelizing Legendre Memory Unit Training. In Marina Meila and Tong Zhang, editors, Proceedings of the 38th International Conference on Machine Learning, volume 139 of Proceedings of Machine Learning Research, 1898–1907. PMLR.  Abstract  PDF  External link

Technical Reports and Preprints

• Narsimha Chilkuri, Eric Hunsberger, Aaron Voelker, Gurshaant Malik, Chris Eliasmith (2021) Language Modeling using LMUs: 10x Better Data Efficiency or Improved Scaling Compared to Transformers. arXiv preprint.  Abstract  PDF  arXiv