Author Archives: Danilo Ardagna

Design of a client-server architecture for the gamification of socio-economic behavioral experiments

Posted on by

Human behavior in different contexts is explored by experimental sociologists and economists through decision-making experiments that reproduce socio-economic scenarios in a stylized way. Traditional experiments are organized in physical labs or through crowd-work platforms and volunteers receive economic incentives to guarantee the validity of their answers.

The general aim of the project is to explore a new way of rewarding participants, by providing them a ludic experience in lieu of the economic reward. As a first step in this direction, the project will develop a client-server architecture to run a virtual game implementing the iterated Prisoner’s Dilemma, a classic game for which many datasets of traditional experiments are available for comparison.

The idea is to exploit the ICT services of Politecnico di Milano to host the server for a mobile app. The hosting services @PoliMi are based on docker containers running on a web server. The developed architecture will be modular and scalable, to serve for future experiments of different nature.

Required skills

– Web application programming and networking

– Docker container technology
– Data base management
– Gitlab version control
Contacts:

Prof. Fabio Dercole (teacher of System’s Theory, DEIB Polimi, fabio.dercole@polimi.it),

Prof. Danilo Ardagna (teacher of Computing Infrastructures, DEIB Polimi, danilo.ardagna@polimi.it)

Machine Learning techniques to Model Data Intensive Application Performance

Posted on by

Nowadays, Big Data are becoming more and more important. Many sectors of our economy are now guided by data-driven decision processes. Spark is becoming the reference framewrok while at the infrastructural layer, cloud computing provides flexible and cost-effective solutions for allocating on-demand large clusters, often based on GPGPUs.  In order to obtain an efficient use of such resources, it is required a model of such systems being at the same time precise and efficient to use.

One common way to model multi-class systems makes use analytical models like queueing networks or Petri nets. However, despite having a great accuracy in performance prediction, their significant computational complexity limits their usage. Machine learning techniques can solve this problem and develop models being accurate and scalable at the same time.

This project involves the development and validation of models for Big Data clusters based on Spark or based on GPGPUs to support deep learning applications training.  The project will develop benchmarking scripts to gather operational data and will compare multiple machine learning algorithms like Support Vector Regression, Linear regression and random forests.