Michael Noukhovitch


Hi, I'm a PhD candidate in artificial intelligence at Mila associated with Université de Montréal supervised by Aaron Courville. I finished my Master's there in 2019 and before that I graduated in 2017 with a Bachelor's in Software Engineering student from the University of Waterloo, during which I also spent an exchange term at Lund University in Sweden.

My research goal is to learn interaction with humans through language but my interests span NLP, reinforcement learning, and more. Recently, I talked about some of my research interests on a podcast!


Simplicial Embeddings in Self-Supervised Learning and Downstream Classification
Samuel Lavoie, Christos Tsirigotis, Max Schwarzer, Ankit Vani, Michael Noukhovitch , Kenji Kawaguchi, Aaron Courville

ICLR 2023 Spotlight (Top 25%)
Code available

Countering Language Drift with KL Regularization
Michael Noukhovitch , Samuel Lavoie, Issam H. Laradji, Douwe Kiela, Florian Strub, Aaron Courville

InterNLP Workshop @ NeurIPS 2022

Competition exacerbates Language Drift
Michael Noukhovitch, Aaron Courville, Issam H. Laradji

Machine Learning and the Evolution of Language Workshop @ JCoLE 2022

Pretraining Representations for Data-Efficient Reinforcement Learning
Max Schwarzer, Nitarshan Rajkumar, Michael Noukhovitch, Ankesh Anand, Laurent Charlin, Devon Hjelm, Philip Bachman, Aaron Courville

NeurIPS 2021
Code available

Emergent Communication under Competition
Michael Noukhovitch, Travis LaCroix, Angeliki Lazaridou, Aaron Courville

AAMAS 2021
Code for Circular Game and Negotiation Game
AAMAS Talk and Slides

Emergence of Communication with Selfish Agents
Michael Noukhovitch, Travis LaCroix, Angeliki Lazaridou, Aaron Courville
EVOLANG 13 Short paper, Oral

Considering Assumptions of Emergent Communication
Michael Noukhovitch
Montreal AI Symposium 2020 Short paper, Poster

Emerging Communication between Competitive Agents
Michael Noukhovitch
Master's Thesis, 2020

Systematic Generalization: What Is Required and Can It Be Learned? Dzmitry Bahdanau, Shikhar Murty, Michael Noukhovitch, Thien Huu Nguyen, Harm de Vries, Aaron Courville

ICLR 2019
Code Available

Selective Emergent Communication with Partially Aligned Agents Michael Noukhovitch, Aaron Courville

Emergent Communication Workshop @ NeurIPS 2018

Oríon: Experiment Version Control for Efficient Hyperparameter Optimization Christos Tsirigotis, Xavier Bouthillier, François Corneau-Tremblay, Peter Henderson, Reyhane Askari, Samuel Lavoie-Marchildon, Tristan Deleu, Dendi Suhubdy, Michael Noukhovitch, Frédéric Bastien, Pascal Lamblin

Workshop on Automatic Machine Learning @ ICML 2018
Code Available

Commonsense mining as knowledge base completion? A study on the impact of novelty Stanisław Jastrzębski, Dzmitry Bahdanau, Seyedarian Hosseini, Michael Noukhovitch, Yoshua Bengio, Jackie Chi Kit Cheung

NAACL 2018 Workshop on New Forms of Generalization in Deep Learning and Natural Language Processing Code Available


Oríon Python

Mila's asynchronous distributed hyperparameter optimization for deep neural networks

Prototypr Python OpenCV Tesseract

Android app that allows you to draw an app mockup on paper, take a picture of it, and quickly turns into a real app interface using React.js, made as part of Hack the North 2015.

Card Android

Android app for creating and sharing electronic business cards using NFC, made as a part of Calhacks 2014

Meta-Reviewer Python Scikit-learn

A machine learning classifier that reviews a Yelp user's reviews. Determines the overall quality of a user's Yelp reviews against training data of 1 million reviews, made as part of Yelp Fall 2014 Hackathon.


ServiceNow Research Visiting Researcher PyTorch HuggingFace

I returned two years later to the great team at ServiceNow Research, fka ElementAI, to work on semi-supervised language as well as continue my work on language models and reinforcement learning. I worked with Issam Laradji and got a better understanding of production requirements and needs for NLP. I was happy to work on different aspects of NLP and we got two neat workshop papers out!

Facebook AI Research Intern PyTorch

I finished off the break between Master's and PhD by working with Douwe Kiela on language pretraining for interaction. I was hoping to work in NYC, which then changed to Menlo Park, and finally ended up being my living room thanks to COVID-19. I still had a good time doing the internship remotely and having awesome video calls with the team in Montreal. I'm looking forward to seeing everyone in person once things settle down. It was also an interesting time to work at Facebook and seeing the discourse from both sides.

ElementAI Research Intern PyTorch HuggingFace

I moved one building over to work with Dzmitry Bahdanau and Harm de Vries at a vibrant Montreal startup! With the grand challenge of learning natural language database interaction, we worked on language pretraining for text-to-SQL. We had an ambitious project that ended up dealing with the fundamental expressivity of SQL and the challenge of zero-shot database adaptation. I enjoyed my time there despite the "trying times" and very fondly remember the team sitting down for lunch together.Also the kombucha.

NextAI Scientist in Residence

I did consulting for NextAI's 2019 Montreal cohort giving machine learning advice to companies working on water data, supply chain management, tracking the dark web, and managing dental information! It was a great experience seeing how ML and data science can be leveraged in creative ways to make real, viable products.

Mila Research Intern Python Theano

I spent my last co-op term working in academia doing deep learning research! I worked on backpropogation through stochastic discrete neurons as applied to GANs and word embeddings coming up with novel ideas and models. I had a great time working at one of the biggest academic deep learning labs in the world and met many smart, insightful grad students. Montreal as a city is also really great! I love the beer, bakeries, and my French really improved!

Google Research Software Engineering Intern Python Tensorflow Blender

I got the amazing opportunity to do deep learning research at one of the top places in the world! I worked on the whole pipeline, from the proposal of an idea, figuring out a specific benchmark, creating a pipeline for generating data, and researching different architectures for the computer vision task. It was a great experience and I learned the basics of applying and researching deep learning, ending up with a pretty good result.
I really enjoyed working with my fantastic mentor, Wei, going to reading groups, listening to tales of the early silicon valley, and meeting fantastic researchers. I delved even more into coffee and learned how to make a pretty decent latte (espresso machine and all), though my latte art needs some work!

Premise Social Capital Fellowship Intern Python Java Scala Spark

I came back to San Francisco to work at a really interesting startup, changing the way we measure the economy (and a bunch of other things!) by using an app for on-the-ground contributors, paying them to collect data, and intelligently processing it to figure out what is happening across the world in real time.
I worked as an everything engineering touching the data analysis pipeline, the android app, the server code, and even getting into some image processing. It was fun working with such a skilled, tight-knit team and I drank a lot of good coffee)

Yelp Software Engineering Intern Python Javascript

I spent a great four months in San Francisco working on biz.yelp.com
I worked on full-stack web development on an agile team, getting my first real taste of Silicon Valley. Then, I dove deep into the code and worked on frontend features like the new landing page and purchasing flow, as well as backend where I internationalized check-in offers. I also discovered the best chocolate.

Watrhub Software Engineering Intern Python ElasticSearch

I worked at an awesome water data startup in downtown Toronto located at CSI
I got to work on our whole backend pipeline, from scraping and web crawling, to unstructured text classification, to text parsing. I delved further into classification and really improved our recall with positive-unlabelled learning, and did some interesting work integrating ElasticSearch with MongoDB for full featured text-search through pdfs
Also the tea was great.

Canadian Government, Data Scientist Python

I worked on some statistical machine learning in fields such as pattern detection and classification, in which I even implemented a novel algorithm!

Contact Me

I am mnoukhov pretty much everywhere, so feel free to email me @gmail or maybe find me on github or linkedin