Post-doctoral fellow on automatic machine learning
Post-doctoral fellow on automatic machine learning
You cannot apply for this job anymore (deadline was 7 Dec 2020).
Browse the current job offers or choose an item in the top navigation above.
Job description
We are seeking a highly creative and motivated post-doctoral fellow to join the Data Mining Group at the Eindhoven University of Technology. The candidate will be working in collaboration with dr. ir. Joaquin Vanschoren, as well as the OpenML core team and the TAILOR European Network of Excellence in AI.
The field of automated machine learning (AutoML) aims to automatically build machine learning models in a data-driven, objective, and automatic way. The user simply provides data, and the AutoML system automatically determines the approach that performs best for this particular application. Although the field is moving fast and substantial progress has been made, current solutions do not yet ensure robust and trustworthy models, and occasionally fail catastrophically.
First, we aim to ensure that AutoML can be used 'in the wild'. You will design methods to automatically handle messy real-world data, integrating methods and developing new methods for feature preprocessing, handling unstructured data, data with many defects, missing data, concept drift, and data wrangling.
Second, AutoML systems should be able to address multiple objectives, automatically determining the best tradeoff between performance and other objectives, including explainability, safety, and fairness.
Finally, AutoML systems should get better over time, producing increasingly better models requiring less and less data. Just like humans can learn new task extremely efficiently by building on a lifetime of experience on other related tasks, research into meta-learning can give us AI systems that will be much more efficient to train (sustainability) and will be more robust since they reuse previously learned patterns and representations. We aim to set up infrastructure (building on OpenML), collect (meta)data and prior trained models, and develop techniques that leverage this metadata to efficiently and automatically design novel open-ended AI systems.
This work will be done in close collaboration with the vibrant AutoML community, including many top AutoML teams at universities and companies.
The field of automated machine learning (AutoML) aims to automatically build machine learning models in a data-driven, objective, and automatic way. The user simply provides data, and the AutoML system automatically determines the approach that performs best for this particular application. Although the field is moving fast and substantial progress has been made, current solutions do not yet ensure robust and trustworthy models, and occasionally fail catastrophically.
First, we aim to ensure that AutoML can be used 'in the wild'. You will design methods to automatically handle messy real-world data, integrating methods and developing new methods for feature preprocessing, handling unstructured data, data with many defects, missing data, concept drift, and data wrangling.
Second, AutoML systems should be able to address multiple objectives, automatically determining the best tradeoff between performance and other objectives, including explainability, safety, and fairness.
Finally, AutoML systems should get better over time, producing increasingly better models requiring less and less data. Just like humans can learn new task extremely efficiently by building on a lifetime of experience on other related tasks, research into meta-learning can give us AI systems that will be much more efficient to train (sustainability) and will be more robust since they reuse previously learned patterns and representations. We aim to set up infrastructure (building on OpenML), collect (meta)data and prior trained models, and develop techniques that leverage this metadata to efficiently and automatically design novel open-ended AI systems.
This work will be done in close collaboration with the vibrant AutoML community, including many top AutoML teams at universities and companies.
Specifications
- max. 38 hours per week
- Eindhoven View on Google Maps
:fill(white)/logos/tue-en-wide_acOLN2b.png)
Eindhoven University of Technology (TU/e)
Requirements
We are looking for a motivated candidate with:
- A Doctorate in Computer Science (or similar);
- Advanced knowledge of machine learning techniques;
- Prior experience in meta-learning or automatic machine learning is an asset;
- Strong mathematical and analytical skills;
- Excellent programming skills. Experience with open source development is an asset;
- Excellent communication skills in spoken and written English;
- Creativity, free thinking, perseverance.
Conditions of employment
We offer you:
- An exciting job in a dynamic work environment
- A full time appointment for 1 year at Eindhoven University of Technology (www.tue.nl/en) with the possibility of an extension with one more year
- The salary is in accordance with the Collective Labour Agreement of the Dutch Universities, increasing from € 2,790 per month initially, up to € 4.402 per month.
- An attractive package of fringe benefits, including end-of-year bonus (8,3% in December), an extra holiday allowance (8% in May), moving expenses and excellent sports facilities.
