Laura Ruis

Artificial Intelligence Student

University College London

GitHub

Twitter

LinkedIn

Welcome

Hi! My name is Laura and I am a PhD student working on language and reinforcement learning in the UCL DARK lab. I'm (very broadly) interested in generalization in machine learning. Finding failure modes of contemporary deep learning methods, and trying to improve them. I want to learn more about the type of inductive biases allowing things like human-like or better systematic generalization and compositionality, and finding ways to learn them from data.

About me

Currently - PhD candidate at University College London supervised by Tim Rocktäschel and Ed Grefenstette

Previously

  • Assistant Research Scientist at New York University in Brenden Lake's Human and Machine Learning Lab

  • Master Thesis Intern at Google in the Perception team working on human perception of audio

  • Research Intern at Facebook Artificial Intelligence Research (FAIR) working on systematic generalization in language

  • Software Engineering Intern at Google in the Assistant team working on automatic text simplification and partly autoregressive transformers

  • Teaching Assistant at University of Amsterdam (UvA) for courses like Data Processing and Natural Language Processing

  • Masters in AI at University of Amsterdam (UvA), graduated cum laude.


News

  • November 2021 -- I wrote a blogpost on Learning in High Dimension Always Amounts to Extrapolation, find it here.


  • December 2020 -- I'll present gSCAN at NeurIPS on Thursday December 10th at 9AM PST in poster session 6. Find the version of the poster with a bit more textual explanation than the one in the proceedings here.


  • September 2020 -- GroundedSCAN got accepted to NeurIPS 2020! The camera-ready version is now available on arXiv.



Research Interests & Experience

Language

Natural language is one of the drivers behind human intelligence, and modeling it is a great challenge.

Experience

    • Experience with structured prediction models for syntactic parsing (blogpost in the making!)

    • Partly autoregressive models for text simplification (proposed in this paper)

    • Grounded language understanding (designed a benchmark)


Cognitive Science

Human and other animal intelligence is remarkable and can be a great inspiration for artificial intelligence.

Experience

    • Designed a benchmark for systematic generalization based on the concept of compositionality at FAIR supervised by Brenden Lake