PhD project on Active Deep Learning for Nano Sensor Systems

Last modified: 
Monday, February 13, 2017 - 15:37

A PhD project in Active Deep Learning for Nano Sensor Systems is available from April 1st at DTU Compute, Technical University of Denmark.

 

Please apply by March 1 via http://www.dtu.dk/english/career/job?id=9f25d4e3-2cd0-463e-b9da-5862b69512da

 

The PhD project is associated with IDUN Sensor research that focuses on development and exploration of nano-mechanical biosensors. Data processing and modelling are indispensable multipurpose tools for sensor development and evaluation and analysis of results in demonstration activities. In sensor development, data modelling tools provide other views and insight into the physical and chemical properties of the sensor as well as sensing principle; hence, improving sensor development in terms of time-use, but also the ability to robustly confirm hypotheses about the sensor’s functionality. In relation to sensor demonstration activities, data modelling is important for obtaining robust sensor performance by suppressing of noise caused by undesired physical and chemical properties of the sensor as well as uncontrollable experimental factors.  

Recent advances within the deep learning field has shown remarkable performance in a great variety of data processing tasks. The PhD project will focus on developing new active learning methods for deep neural network models. Such methods can provide optimal experimental design and hypothesis testing for sensor development, and further reduce the need for user labels in connection with demonstration of detection and predictive sensing capabilities. The methodological research relates to, and will leverage from, current advances in Bayesian optimization; one-shot-learning; generative adversarial networks; and users-in-the-loop models, where the user is the sensor developer and/or an end-user providing labeled information. 

 

Requirements
Candidates must have a master degree in either machine learning, computational science and engineering, applied mathematics, engineering, or equivalent academic qualifications. Preference will be given to candidates who can document knowledge in machine learning, neural networks, and sensor information processing and in addition have a background and experience with Bayesian statistics, experimental design and sensor systems. Furthermore, good command of the English language is essential. 

 

 

Further Information
Further information concerning the project can be obtained from Professor Jan Larsen, +45 4525 3923, janla@dtu.dk and Senior Researcher Tommy S. Alstrøm, +45 4525 3431, tsal@dtu.dk

 

Best regards, Jan Larsen