Three (3) PhD Positions in Information & Communication Technologies, University of Luxembourg, Luxembourg (2015-2016)

nT is a recently formed centre carrying out interdisciplinary research in secure, reliable and trustworthy ICT (Information and Communication Technologies) systems and services, often in collaboration with industrial, governmental or international partners (www.securityandtrust.lu). The research work will be performed within a young and enthusiastic Big Data working group, established within the larger NetLab research group: http://wwwen.uni.lu/snt/research/netlab. We are looking for a PhD candidate strongly motivated in working with an industry leading company in this area and to be willing in showing leadership, commitment and excellent analytic capabilities.

PhD candidate in Self-learning Predictive Algorithms: from design to scalable implementation (M/F)

Ref: I2R-DIR-PIC-09SCNI
Fixed Term Contract 3 year (CDD), full-time (40 hrs/week)
Number of positions: 3

The research wok is done in the context of a starting collaboration with a partnering company Ola Mobile, in the area self-learning predictive algorithms for effective Mobile Gaming Monetization. The major topic concerns the development of self-learning algorithms capable to perform a permanent learning of how mobile devices behave with respect to a large set of input factors. The factors range from a broad scope of relevant features, that can be static (device, brand, location) but also behavioural related. Examples of the latter can be in-application purchases effectuated during the process. This proposed work envisages also the development of scalable approaches to implement such predictive methods. More specifically, it will address how such predictive algorithms can be designed in order to leverage the current state of the art architectures for Big Data (Hadoop for batch processing, STORM/SAMZA for online streaming analytics)
The tasks of the PhD candidate will consist into performing research analysis as well as designing and evaluating proposed solutions through real implementation in partnership with an industry partner. The research associate will have the opportunity to present his or her results to the international community by participating to worldwide scientific events. More specifically, the work will address the following issues:

- State of the art research in existing conceptual solutions for self-learning predictive algorithms
- Practical assessment of open source and commercially (if needed) tools required for the concrete project
- Exploratory data analysis on the dataset in order to identify the relationships among the different features (PCA, factor analysis, correlations, etc).
- Identification of relevant features and supporting data-structures for efficient representation and I/O bound operations on the feature set.
- Design of a Deep-Learning (neural network) architecture for self-learning.
- Scalable prediction engine based on a Lambda architecture (Storm/SAMZA, Spark/Shark Hadoop) jointly with a distributed Machine Learning algorithm over a Trident topology that integrates the previous Deep learning solution.
- Comparisons with alternative solutions (A/B Testing, Ensemble learning) as well as performance results (number of features/classifications/seconds, memory and storage requirements)
- Realistic field trial and model updating to match with real-world requirements and experience.

Further details:
http://academicpositions.eu