Nayan Saxena

Machine Intelligence + Statistics

Get Started

Here is a video of me generating the cover image

About Me

I am a recent graduate from the University of Toronto where I completed an undergraduate degree in Statistics, Computer Science and Mathematics. Furthermore, I am an independent machine learning researcher at the ML Collective where I have worked on projects pertaining to Neural Architecture Search, low-resource language translation and Federated Learning. I have also worked as an ML Engineer at one of Canada's fastest growing startups, WOMBO.ai, where I focused on AI generated art. Furthermore, I formerly interned as a data scientist with the artificial intelligence team at Bell Canada and have written several papers on topics including Neural Architecture Search, Dueling Bandit Algorithms, Item Response Theory & Gaussian Processes presented at top machine learning venues like AAAI, ICML, NeurIPS, EDM, IEEE-ICALT, and ICCM. My other interdisciplinary projects span topics such as multi-armed bandit algorithms, clinical trials, and population dynamics during COVID-19.

Teaching Assistant:

STA130: Introduction to Statistical Reasoning and Data Science (Winter 2022)
STA302/1001: Methods of Data Analysis I (Fall 2021)
STA247: Probability with Computer Applications (Fall 2021)
STA130: Introduction to Statistical Reasoning and Data Science (Fall 2020)

Reviewer / Program Committee:

Other Affiliations : FOR.ai , UofT Artificial Intelligence Group & UofT Machine Intelligence Student Team

Publications

NSL in Practice Sign-to-Speech Model for Sign Language Understanding: A Case Study of Nigerian Sign Language
Steven Kolawole, Opeyemi Osakuade, Nayan Saxena & Babatunde Olorisade
In Proceedings of the 31st International Joint Conference on Artificial Intelligence, Vienna, Austria, 2022
Workshop on Machine Learning for the Developing World, 35th Conference on Neural Information Processing Systems (NeurIPS), 2021
Preprint | Oral Talk | Code

Neural Architecture with Few Bad Operations Towards One Shot Search Space Poisoning in Neural Architecture Search (Student Abstract)
Nayan Saxena, Robert Wu & Rohan Jain
In Proceedings of the 36th AAAI Conference on Artificial Intelligence, Vancouver, BC,Canada, 2022
Preprint | Code | Poster

Neural Architecture Type System NeuralArTS: Structuring Neural Architecture Search with Type Theory (Student Abstract)
Robert Wu, Nayan Saxena & Rohan Jain
In Proceedings of the 36th AAAI Conference on Artificial Intelligence, Vancouver, BC,Canada, 2022
Preprint | Code | Poster

Sampling Process Picture Dynamic Strategy Selection in Active Function Learning
Nayan Saxena, Rebekah A. Gelpi, Daphna Buchsbaum & Chris Lucas
Abstract in proceedings of the 44th Annual Meeting of the Cognitive Science Society, Toronto, Canada, 2022
Poster | Abstract | Code

Gaussian Process Picture Sampling Heuristics for Active Function Learning
Rebekah A. Gelpi, Nayan Saxena, George Lifchits, Daphna Buchsbaum & Chris Lucas
In Proceedings of the 19th International Conference on Cognitive Modelling, 2021
Abstract in proceedings of the 43rd Annual Meeting of the Cognitive Science Society, Vienna, Austria, 2021
Paper | Poster | Talk

sparse_matrix_student IRT++: Improving Student Response Prediction With Gaussian Initialisation and Other Modifications
Nayan Saxena, Varun Lodaya & Trisha Thakur
In Proceedings of the 21st IEEE International Conference on Advanced Learning Technologies, Tartu, Estonia, 2021
Paper | Code

Neural Architecture Search (NAS) Search Space Poisoning the Search Space in Neural Architecture Search
Robert Wu*, Nayan Saxena* & Rohan Jain* (*equal contribution)
Workshop on Adversarial Machine Learning, 38th International Conference on Machine Learning, 2021
Long Preprint | Workshop Paper | Poster | Code

Bandit Optimisation Statistical Consequences of Dueling Bandits
Nayan Saxena, Pan Chen & Emmy Liu
Workshop on Reinforcement Learning for Education, 14th International Conference on Educational Data Mining , Paris, France, 2021
Paper | Preprint | Spotlight Talk | Code

texas_social_distancing Political and Socioeconomic Influences on Social Distancing Behaviour in the United States
Liam Keating, Nayan Saxena, Emma Cooper, Jordan Tirico, Daniel Khain & Daphne Imahori
SocArXiv, 2020 (doi:10.31235/osf.io/emrtj)
Paper

Projects

COVID-19 & Crime Dashboard

A dashboard focused on COVID-19 and US Crime data that makes future predictions on crime rates across Seven US Cities using Vector Auto Regressive Models and provide crime estimates before and after onset of the pandemic to gain insight into the effect of self-isolation measures on human behavior. This project won the Gold Prize for Best Dashboard based presentation out of 42 competing teams at the American Statistical Association Datafest.

See Dashboard

DOT.

A binary classifier between open and closed eyes that seamlessly converts eye-blinking data into Morse code and finally speech, trained and presented at Hack The North 2019 (University of Waterloo). The project uses Keras with a Tensorflow backend to convert real-time blinking data to speech by translating data input into Morse code and further decrypting data into audio output and keystrokes.