Prominent AI conference features 9 papers & posters from CU Boulder
Master's student Tyler Scott gives a presentation on his team's work during NeurIPS.
The Department of Computer Science’s AI chops were on full display this month at NeurIPS, the year’s biggest conference on machine learning and artificial intelligence.
Professor Mike Mozer serves as secretary on the conference’s board, while Associate Professor Claire Monteleoni was an area chair on the program committee. Monteleoni also serves on the senior advisory council of the Workshop for Women in Machine Learning, which takes place annually at NeurIPS.
In addition to helping plan the conference, which sold out its 8,000 seats within hours and had 150 industry sponsors, Mozer and Monteleoni’s teams also had several papers and posters accepted to the conference.
Monteleoni’s postdoctoral researcher, Sophie Giffard-Roisin, presented a poster on Deep Learning for Hurricane Track Forecasting from Aligned Spatio-temporal Climate Datasets in the spatiotemporal domain workshop. Mozer’s team – including CU Boulder students Karl Ridgeway and Tyler Scott – presented three papers:
- Sparse Attentive Backtracking: Temporal Credit Assignment Through Reminding
- Adapted Deep Embeddings: A Synthesis of Methods for k-Shot Inductive Transfer Learning
- Learning Deep Disentangled Embeddings With the F-Statistic Loss
Assistant Professor Rafael Frongillo and PhD student Jessie Finnochiaro also had a spotlight paper, Convex Elicitation of Continuous Properties, in the main NeurIPS conference, while visiting scholar Vincent Cohen-Addad presented on Clustering Redemption–Beyond the Impossibility of Kleinberg’s Axioms.
Bo Waggoner, who will join the department as an assistant professor in 2019, had three papers in the conference:
- A Smoothed Analysis of the Greedy Algorithm for the Linear Contextual Bandit Problem
- Local Differential Privacy for Evolving Data
- Bounded-Loss Private Prediction Markets (with Frongillo)
Stephen Becker of applied math also had a paper, Low-Rank Tucker Decomposition of Large Tensors Using TensorSketch, accepted to the conference.
Photo at right: Rafael Frongillo, Bo Waggoner and Jessie Finnochiaro of CU Boulder with Daniel Mutembesa of Makerere University in Uganda.