The International Conference on Machine Learning (ICML) will be taking place Sunday, July 18 to Saturday, July 24 as a virtual event this year. ICML is one of the most relevant machine learning conferences in the world, known for presenting and publishing cutting-edge research and applications on all aspects of machine learning including: artificial intelligence, deep learning, machine vision, computational biology, speech recognition, robotics and more. SigOpt has been an active participant of ICML since 2016. This year, SigOpt will be presenting our research on a new methodology to power model tuning:
Beyond the Pareto Efficient Frontier: Constraint Active Search for Multiobjective Experimental Design
by the members of our research team, Gustavo Malkomes, Harvey Cheng, Eric Lee, and Michael McCourt. Stay tuned for an upcoming blog post about this paper, and how this method can power teams to learn more about their problem space and metric tradeoffs. There will also be a 5 minute presentation and a poster session about this work.
Spotlight Presentation: Bayesian Learning 1
Thu 22 July 6:20 – 6:25 AM PDT
Poster Session 5
Thu 22 July 9:00 – 11:00 AM PDT
Interview: AutoML Research in Active Search and Multiobjective Experimental Design
TWIML does a great job of covering the major research conferences every year, including discussions with authors of some of the most exciting papers. This year, podcast and YouTube host Sam Charrington will sit down with Gustavo to discuss his research that led to the paper, Beyond the Pareto Efficient Frontier: Constraint Active Search for Multiobjective Experimental Design. During the discussion, they’ll discuss Active Learning, how it complements AutoML techniques like Bayesian optimization, novel methods for applying this method to multiobjective experimental design, and how this type of research can be implemented to help teams more robustly explore their modeling problem space.
Connect with us at the Intel Booth
Interested in learning more about SigOpt? Find us at the Intel booth to meet our team. We will share relevant content relating to our product, model training, hyperparameter optimization, Bayesian optimization and more.
Try SigOpt for free
SigOpt enables any modeler to track their training runs and tuning jobs, visualize and compare training runs with customizable visualizations to gain intuition on your models, and easily share projects to collaborate with your team. SigOpt is fully agnostic, compatible with any modeling framework, compute stack or coding environment. You can get free access to the product here.