NeurIPS 2019 Reproducibility Challenge- Kernel-Based Approaches for Sequence Modeling: Connections to Neural Methods

Palak Goenka, Kartikey Pandey, Ashutosh Bhushan Bharambe, Subham Sahoo

Abstract: Motivated by the importance of kernel-machines for the development of Deep Learning models, we have represented an extensive study of the work done in the paper Kernel-Based Approaches for Sequence Modeling: Connections to Neural Methods as a part of NeurIPS Reproducibility Challenge 2020. In this report, we have tried to judge the reproducibility of the original paper by comparing our results with the ones reported by the authors. Along with every minute explanation required for the experimentation defined in the original paper, we have also tried to provide insights for the incorporation of various training techniques mentioned by the authors. Our complete codebase is available at \href{https://github.com/palakg11/KASM-Pytorch}{\underline{{\textcolor{blue}{GitHub}}}}.

Track: Replicability

NeurIPS Paper Id: https://openreview.net/forum?id=S1gYwVHxUH