Bachelor Thesis proposal

Learning from neural or stochastic networks?

Neural networks represent the traditional framework for modelling the learning capability of a machine. They have been succesfully applied for recognizing characters or spoken languages. However, other approaches modelling the human brain have been proposed in literature, specifically the G-networks. These, similarly to the standard neural networks, can be trained using a data set but, differently from their predecessors, they have a stochastic nature. This thesis aims at implementing the two models and compare them under different scenarios exposing the strengths and weaknesses of both approaches.


Veneto eGovernment Lab – Smart eGovernment Dashboard (SmeD)

veg-lab-text  SmeD

Context: project Smart eGovernment Dashboard (SmeD) with the Regione Veneto local government for the objective measurement of the effectiveness of the public interventions in e-Government & ICT. Creation of a virtual dashboard in order to represent the relevant indicators.

What’s SmeD?: il sistema SmeD potrà consentire alla Regione Veneto l’attivazione di un monitoraggio continuo del territorio tramite un’appropriata qualificazione e quantificazione di indicatori socio-economici, ambientali e di mobilità, agganciati su una dimensione predefinita (territoriale, merceologica, sociale, …). La raccolta degli indicatori rilevanti potrà avvenire per via automatica (via webbots / scrapers) e semi-automatica (via estrattori / wrappers), con la possibilità di completare i dati ove necessario con campagne di indagine focalizzate.

SmeD 1: Client bot engine: webbots, spiders & crawlers experimentation

Activity: consolidation and evolution of the “webbot client engine”, i.e. of the (java multithreaded) support platform where to activate / schedule / experiment webbots & spiders; design and realization of webbots / spiders (java) able to navigate independently the web collecting data, testing graph-walking spiders for social media, implementation and testing of text-searching spiders for correlation analysis of binary / ternary keywords.


SmeD 2: SmeD Server design: requirements & tests

Activity: design, construction and testing of a (java / servlets multithreading) “webbot server engine” capable of supporting management services, add, remove, scheduled execution of webbots / spiders. For this system will be evaluated / tested the possibility of activation of infrastructure “cloud computing” existing / available (SaaS / PaaS). Design, implementation and testing of interfaces, management / monitoring (a “dashboard”) to interact with the server engine will be researched.


SmeD 3: Cloud computing architecture(s) for SmeD: Google PaaS development environment

Activity: the goal is to evaluate / test the Google App Engine PaaS for the development and test of the SmeD Server component. The Google Cloud Engine IaaS should also be evaluated in terms of costs/parameters as a production environment for SmeD Server webbot engine and for the the deployment of the underlying open source search engine.


SmeD 4: Open Source Search Engine for SmeD: Apache Lucene Solr

Activity: deployment and testing of Apache Lucene Solr, a search engine to be activated on a virtual server IaaS as a “base” on which to perform queries KPIs. The search engine will be activated “in the cloud” and its implementation will be evaluated (and measured in terms of thorughput / band consumption) in order to estimate its performance & costs in a production system.


SmeD 5: SmeD Mobile interfaces: analysis of technologies & representations

Activity: design, implementation and testing of mobile dashboard interfaces / apps for tablets / smartphones able to interact with the “webbots server engine”. Interfaces and dashboards should be easy and immediate and designed for subjects such as mayors / city councilors who, with little time, need to know the situation, assess their effects and rethink existing policies without having to deal with too much technological complexity. Also, overview of the more popular APIs for representations: Google motion chart APIs for animated charts, Google fusion tables for georeferentiated reprsesentations.