Deep Ensemble Learning

Ensembling is a conventional machine learning technique that combines several base models in order to produce one optimal predictive model. There are many motivations why we might be interested in this technique.

In this project we aim to find better ways for ensembling deep neural networks with direct implications on resource constrained devices.

Download as PDF

Student Target Groups:

  • Students in ICE and Computer Science.

Thesis Type:

  • Master Project / Master Thesis

Goal and Tasks:

  • Literature review on deep learning loss landscape and its application in the context of deep ensembles.
  • Train different neural networks from different initializations and form the conventional ensemble.
  • Investigate the functional properties of the ensemble.
  • Implement new ways for forming the ensemble.
  • Summary of the results in a written report, oral presentation.

Recommended Prior Knowledge:

  • A good knowledge of neural networks and interest in understanding why they work well!
  • Programming skills in Python.
  • Prior experience in deep learning frameworks is desirable (preferably PyTorch)

Used Tools / Equipment:

  • A laptop (GPU infrastructure will be provided)
  • Your talent (very important!).


  • a.s.a.p.