Continuous Learning of Deep Neural Networks

  • Sponsored by: PreciBake GmbH
  • Project Lead: Dr. Ricardo Acevedo Cabra
  • Scientific Lead: Mathias Sundholm
  • Term: Summer Semester 2019

Results of this project are explained in detail in the final documentation and presentation.

PreciBake, is a small Munich based company developing AI solutions for food and baking industry. Our AI team is continuously working on developing and improving our ML algorithms for tasks such as image classification.

Most machine learning system today relies on supervised training of models. Updating and improving a trained model is a lengthy manual process that requires cleaning and labeling of new acquired data, and repeated retraining and testing of the model. Maintaining large, clean and uniform data sets becomes increasingly time consuming and requires often expert knowledge.

With recent advancement in transfer learning, semi supervised learning and reinforcement learning a model could learn to utilize new unlabeled data in order to automatically improve itself over time with only sparse or no supervision of human experts.

In this project you will explore the most recent advances in deep learning and reinforcement learning research, in order to design an AI system that automatically improve itself over time as new data becomes available.