View Uppdrag

Uppdrag

Rubrik

Organisation/Företag

Plats

Uppdrag

Beskrivning

Background: The technological advancements of the past decade have enabled the rise of decentralized learning as an effective and privacy-preserving paradigm. In decentralized learning, a set of interconnected agents gather data and use it to cooperatively train a model, without the need to store and process all data in a single location. However, these devices often have limited resources, e.g. in terms of computational power. Therefore, we need to design cooperative learning techniques that exploit these resources efficiently. In particular, the aim is to improve the performance of the learned model (for example accuracy in a classification task) by accurately selecting the hyperparameters. The challenge is that hyperparameter selection, needs to be performed in a cooperative fashion by the agents.

Objective: The project focuses on the development of hyperparameters selection/optimization techniques for cooperative learning. The roadmap of the project is as follows: 1) guided literature review, 2) design of novel hyperparameter optimization algorithms, 3) application to learning (possible topics: healthcare, robotics, transportation), 4) numerical evaluations and comparisons. The project has the potential to lead to conference and journal publications in top venues.

We are looking for 1 student to join this project. Requirements are: some background on machine learning, good programming skills (Python), and optionally background on optimization.

The start date is flexible, and the project will take around 20 weeks.

The project will be conducted in the Division of Decision and Control Systems, with flexible remote work, under the supervision of Dr. Nicola Bastianello (nicolba@kth.se) and Dr. Andrea Martin.

Det finns inga poster att visa.

Du har inte behörighet att visa de här posterna.

Fel när begäran slutfördes.

Läser in ...

Sista ansökningsdatum

Publiceringsdatum

Omfattning

Applikationskanal

E-post för ansökan

Appens URL

Ansökningshandlingar

CV, academic transcripts, motivation letter (optional)