Postdoctoral position in mobile activity recognition incl. deep learning

The Wearable Computing Group at the University of Sussex is looking for a postdoctoral research associate to work on a research-intensive project aiming at recognising the manner in which users of mobile phone move about in their daily life from mobile phone sensors. This project is funded by a large multinational active in the mobile device and telecommunication sector.
Candidates should have a PhD in Computer Science, Mathematics, Electrical Engineering or related field. An established expertise in Activity Recognition and Wearable/Mobile Computing is desired with an ability to think of innovative solutions to activity recognition.
Specific skills sought after include:
* Experience in data collection, including defining protocols, handling annotation challenges and doing day to day supervision of the people that will contribute to the data collection.
* Experience in machine learning including feature extraction and selection and traditional machine learning techniques (e.g. SVM).
* Experience in deep learning techniques (e.g. convnets, LSTM).
* Ability to organise and work with large datasets including managing annotations
* Experience with common data science toolkits, such as Weka, NumPy, MatLab, including deep learning toolkits including Theano, TensorFlow, Lasagne, Keras, etc.
* Proficient in Android programming to adapt our data collection app and able to manage the Linux server receiving collected data.
* Strong interest in the combination of theoretical and experimental research.
* Communicative, enthusiastic and good a team player.
*Project objectives*
The overall aim of the project is to develop new techniques to recognise how people move about in their daily life from the sensors in mobile phones.
As part of this, the project will require first to establish a reference dataset. Your first objective will be to contribute to collecting a reference dataset by supervising about 3 project assistants who will be tasked with carrying mobile phones and a body-worn camera and engaging in 8 activity classes. You will define the data collection protocol and implement strategies to ensure data quality including re-annotation by the project assistants. A second objective will be to define a baseline recognition system using traditional machine learning techniques (e.g. SVM). As a third objective you will explore new approaches using deep learning and recurrent networks (e.g. convnets and LSTM) and will contribute to innovation ideas which may result in patents.
*Advantages and career development*
This position is ideally suited for somebody who wants to demonstrate his/her expertise in a highly focused project that will deliver high impact publications, public dataset and possibly patents. This is a career enhancer that will allow the candidate to gain international visibility and collaborate with a large multinational active in the mobile device sector.
The candidate will be supported in applying for grants to support his/her further career development.
Applications should be accompanied by a full CV and a statement of how you envisage your role. The position is initially for 9 months but can be extended up to 12 months, depending on the starting date. Contact Dr Daniel Roggen: D.Roggen-AT-sussex.ac.uk for clarification and informal inquiries.
The full job advert is here: http://www.sussex.ac.uk/aboutus/jobs/1681
Closing date for applications: 8 March 2017