PhD position - KU Leuven - Scalability of Data Analytics
KU Leuven - department of Electrical Engineering collaborates with Philips (Netherlands) and the University of Macerata (Italy) in the MarieSklodowska-Curie project HEART (MSCA-ITN Industrial Doctorate- Grant Agreement No. 766139,http://www.heart-itn.eu). HEART investigates the design and development of a "HEalth related Activity Recognition system based on the Internet-of-Things (loT); HEART is an international, intersectoral and interdisciplinary training program providing PhD Fellowships to 6 Early Stage Researchers (ESRs). This ESR will focus on the development of machine learning and advanced statistical algorithms for scalable data analytics of loT health data.
You will be working with the research groups of e-Media (www.kuleuven.be/eMedia) situated at group-T Leuven, and Advise (www.kuleuven.be/advise) of the faculty engineering technology. The e‐Media Lab investigates, develops and implements novel techniques to enhance the human condition with electronic media. Applications are found in the domain of health care, learning and entertainment. The eMedia research group has in the past developed systems for activity recognition based on audio sensors, video sensors, accelerometers, home automation sensors and occupancy sensors. The group has a strong expertise in the development and
implementation of machine learning methods, data visualization and engagement tools to interact with the end‐user. Advise ‐ Advanced Integrated Sensing lab unites a broad experience in both software (real‐time processing and statistical analysis of large multimodal datasets) and hardware (design and testing of both PCB and integrated circuit implementations). This unique combination of expertise enables the development of integrated sensing and communication systems for different applications including health monitoring, quality control monitoring, audio monitoring among others. You will work on the topic “machine learning for scalable data analytics”.
1) ESR2 will investigate the classification of activities in a subject specific setting based on multimodal
datasets consisting of motion, & vital sign modalities. A focus will be made on activities that play a
fundamental role in the follow-up of lifestyle of people, e.g. cooking, washing, eating and doing sports.
Such activities are composed of a collection of primitive activities and are termed complex activities.
The follow-up of these activities may demand additional contextual information like time of the day,
spatial location, or interaction with other people and objects. Also, specific conditions and typical
situations in China are considered, in the consumer and market perspective.
2) ESR2 will investigate the difficulty of scalability of recognition and classification tasks. The inter
variability between different persons in executing complex activities demands subject-specific models.
However the training of a personal activity model requires a tremendous amount of labelled data,
which is infeasible in practice as this is not available. Therefore we will aim to develop machine learning
approaches that can mitigate the problem of limited availability of labelled data. Transfer-based and
multi-task approaches will play a crucial role in this.
3) ESR2 will support the data-acquisition of a multimodal database consisting of motion & vital sign
modalities. An efficient approach to come to a high-quality labelled set is investigated.
Algorithms that are able to learn in an online and incremental fashion a person specific model of complex activities.
A library of software code for the use of such models, calibrated on preliminary and historical data
INDICATIVE PLANNED SECONDMENTS - Institution, place and timing expressed in contract month (M)
- University of Macerata (Macerata, Italy) - M4
- Philips Electronics Nederland B.V. (Eindhoven, The Netherlands) - M5-M7
- Philips Electronics Nederland B.V. (Eindhoven, The Netherlands) - M12-M14
- Philips Research China (Shanghai, China) - M15-19
- University of Macerata (Macerata, Italy) - M25
- Philips Electronics Nederland B.V. (Eindhoven, The Netherlands) - M26
- Philips Electronics Nederland B.V. (Eindhoven, The Netherlands) - M27-28; 29-33
THE COMPLETE CALL FOR APPLICATION AND LINK TO APPLY CAN BE FOUND AT:
This job comes from a partnership with Science Magazine and