Mauriana Pesaresi Seminars 2017 Speaking Calendar
This is the page for the Mauriana Pesaresi Seminar of 2017.
- 14 March 2017
Speaker: Roberto Pellungrini
Title: A Data Mining Approach to Asses Privacy Risk in Human Mobility Data
Abstract: Human mobility data are an important proxy to understand human mobility dynamics, develop analytical
services and design mathematical models for simulation and what-if analysis. Unfortunately mobility data
are very sensitive since they may enable the re-identification of individuals in a database. Frame-
works for privacy risk assessment give to the data providers the tools to control and mitigate privacy risks,
but they suffer two main shortcomings: (i) they have a high computational complexity; (ii) the privacy risk
must be re-computed every time new data records become available and for every selection of individuals,
geographic areas or time windows. A possible solution is to train classifiers to capture the relation between individual
mobility patterns and the level of privacy risk of individuals.
- 28 March 2017
Speaker: Lucia Nasti
Title: Modelling and Simulations of Dopaminergic System. The case of internet addiction
Abstract: The aim of this project is to show how we can improve our knowledge of neurological structures, using the bioinformatics.
We have developed a basic mathematical model of Dopaminergic system, implementing the theory of Timed and Hybrid Automata.
In this way, we have studied the reaction of the system to different types of pulse: constant in time, at increasing intensity and at a close frequency.
In particular, two kinds of simulations are made:
the first to analyse the relationship between user and technological object and the second to analyse the relationship between users in a computer-mediated communication.
The results obtained are the starting point to develop probabilistic Hybrid automata.
- 4 April 2017
Speaker: Athanasios Rizos
Title: Usage Control in Internet of Things
Abstract: Usage Control is an extension of Access Control.
Although Access Control evaluates attributes only once, before the start of a session, Usage Control (UCON) can deal with them if theychange during this session.
Beyond Access Control, UCON provides two main novelties which are continuity of control and on mutability of attributes that might cause policy revaluation which might lead to revocation.
Security and privacy are important requirements for IoT due to the inherent heterogeneity of the Internet connected objects and the ability to monitor and control physical objects.
However,proprietary security solutions do not help in formulating a coherent security vision to enable IoT devices to securely communicate with each other in an interoperable manner.
One of the most popular application layer protocols used for information sharing in IoT isMessage Queue Telemetry Transport (MQTT) which is a lightweight broker-based Publish/Subscribe messaging protocol standardized in 2013 by OASIS.
My main goal is to integrate Usage Control with IoT protocols and especially with MQTT to achieve secure data sharing.
Furthermore, I have created a survey towards all famous IoT application layer protocols such as CoAP, XMPP, AllJoyn, etc. to declare why MQTT is the most appropriate to collaborate with UCON.
- 11 April 2017
Speaker: Stefano Forti
Title: QoS-aware Deployment of IoT Applications to Fog infrastructures
Abstract: Due to the volume, variety and velocity of data generated by IoT sensors and actuators, the Cloud cannot fully support IoT applications that must meet compelling latency or bandwidth.
Fog computing aims at extending the Cloud by selectively moving compute, storage, communication, control and decision making closer to the edge of the network where data is produced.
Deploying multi-component applications to Fog/Cloud nodes in a QoS- and context-aware manner is challenging due to the heterogeneity and scale of Fog infrastructures.
Application components must be provided with the software and hardware capabilities they need.
Communication links that support interactions between components must meet certain QoS (latency and bandwidth).
As a first step to tackle this challenge, we presented FogTorchΠ, a prototype capable of determining deployments of multi-component applications to Fog infrastructures, which fulfil software, hardware and QoS requirements.
FogTorchΠ classifies eligible deployments in terms of metrics which can support IT experts in deciding how to deploy their multi-component applications.
- 20 April 2017
Speaker: Davide Neri
Title: Introduction to Operating System virtualization (and Docker)
Abstract: Virtualization is used in data center and cloud environments to decouple applications from the hardware they run on.
Hardware virtualization and operating system level virtualization are two prominent technologies that enable this.
Containers, which use OS virtualization, have recently surged in interest and deployment.
A notable example among the newly proposed OS virtualization technologies is Docker.
Docker is a platform that allows to package any application, together with its dependencies, into isolated virtual containers. Docker containers are
lightweight and portable, as they can be quickly deployed and executed on any host where Docker is installed.
- 2 May 2017
Speaker: Parvaneh Parvin
Title: Detecting Anomalous Behaviour in Ambient Assisted Living Scenarios
Abstract: Current technology offers devices and applications to detect a wide range of environmental and user-related parameters. Monitoring such parameters allows us to build knowledge about the context around the elderly.
However, tools able to integrate with real-time data coming from all the potential sensors are missing, as well as tools for analyzing such data and acting consequently when certain conditions are met.
It is fundamental that data analysis, both for short term (alerts) and long term (behavioural) purposes, can be performed taking into account the specificities of the context where the system is used.
In order to fit the particular needs, requirements and routines/tasks of the considered user, it is required that the rules for triggering adaptations and alerts be customizable according to the elderly's habits. In the same way, the approach to behaviour analysis should be customizable based on the local uses.
For instance, different people may have different habits by the time they get up or go to sleep (even at a more general level differences exist in the number and time of meals between various countries, due to traditions, weather, hours of light, etc.).
To this aim, the architectural characteristics of the monitoring approach should be clearly defined in order to show how our solution is able to support the requirements identified in the project.
- 9 May 2017
Speaker: Marco Meoni
Title: Predicting Dataset Popularity for CMS Big Data
Abstract: The Compact Muon Solenoid (CMS) experiment at the European Organization for Nuclear Research (CERN) deploys its PB-size data collection, simulation and analysis activities on a distributed computing infrastructure involving more than 50 sites worldwide.
This work investigates how to leverage data mining on this huge amount of dataset access logs and discover patterns and correlations useful to enhance the overall efficiency of the distributed infrastructure.
In particular we propose a scalable pipeline of Spark components whose goal is collecting from different sites the dataset access logs, organizing them into weekly snapshots, and training predictive models able to accurately forecast which datasets will become popular over time.
Dataset popularity predictions are then exploited within a novel data caching policy, called Popularity Prediction Caching (PPC).
We evaluate the performance of PPC against popular caching policy baselines like LRU and its variations.
The experiments conducted on large traces of real dataset accesses show that PPC outperforms LRU reducing the number of cache misses up to 20%.
- 16 May 2017
Speaker: Benjamin Paaßen (CITEC Center of Excellence, Bielefeld University)
Title: Distance-Based Machine Learning
Abstract: Most classic machine learning approaches rely on a representation of the data in terms of numerical features, such as brightness values for every pixel in an image or frequencies of words in a text document.
However, in many applications, such a vectorial representation may be hard to obtain and/or counter-intuitive, especially when structured data such as sequences, trees of graphs are concerned.
An alternative is a representation in terms of distances. Intuitively, it is easier to state that two objects are similar to each other than to verbalize what their precise features are.
Conversely, if we know that two objects are similar to each other, we can expect them to have similar features, for example, to belong to the same class of objects. This basic assumption enables us to develop machine learning methods based on distances.
This talk will cover how to quantify distances effectively, as well as how to approach classic machine learning problems using distance-based algorithms.
For participants without machine learning background, the talk should provide an intuitive access to the topic and some key algorithms.
Participants with experience in machine learning will be presented with a new and uncommon perspective on the field.
- 23 May 2017
Speaker: Mattia Papini
Title: Urban Tribes: the usage of Stigmergy in City Science
Abstract: The City Science field of research has known an unprecedented development during the last decade.
This huge growth is mostly due to the recent increase of Big Data available about cities and citizens that has led to an unprecedented comprehension of the social dynamics that rule the cities we live in.
The study of these new Big Data has revealed, in particular, that one could distinguish several distinct groups of citizens, or Tribes, in cities.
Tribes are huge entities: the way they act around cities have, intuitively, a huge impact on how these cities’ services are accessed and, indirectly, on their economic performances.
In this context, Computational Social Science is trying to deeply analyse these social relations among humans in urban environments, using new methodologies capable to reveal new insights into our society.
One of these newest methodology is Stigmergy, a concept borrowed by Entomology, that is capable to analyse the way people move across space and time.
The aim of this talk is to introduce the concepts of Tribe and Stigmergy in the City Science context.
In particular, the potential use of Stigmergy in such applications will be assessed by discussing some first results achieved on a Living Lab dataset and the Tribes detected in it.
- 30 May 2017
Speaker: Amaury Trujillo
Title: Fuzzy just-in-time adaptive interventions for self-care apps
Abstract: Health promotion has traditionally been vested in healthcare professionals, but advances in mobile technology, the ever-increasing burden of healthcare systems, and the ideology of patient empowerment have provoked a shift toward personal self-care.
Moreover, due to the prevalence of smartphones and the availability of affordable wearable technology there has been an explosion of health and fitness apps for self-care.
However, most of these apps, even those aimed at particular conditions, are merely trackers that do not adapt to users' health status and behavior, which leads to poor adherence.
These health aspects are inherently vague and difficult to model, and often it is only possible to rely on the expertise of health-related professionals, as there are no reliable datasets nor models for non-clinical contexts.
In this talk, we will see how fuzzy logic can deal with such vagueness, and how just-in-time adaptive interventions, via fuzzy rule-based systems, could be used for self-care apps that are persuasive and useful from the get-go.
- 1 June 2017
Speaker: Giovanna Rosone
Title: Combinatorics On Words and Burrows-Wheeler Transform
Abstract: Words (strings of symbols) have an important role in computer processing. Indeed, each bit of data processed by a computer is a string, and most of computer software use algorithms on strings.
Combinatorics on Words belongs to discrete mathematics and theoretical computer science and its main goal is the study of general properties of words.
The motivation and applications of this theory are manyfold in several areas such as data compression, sequence analysis, computer graphics, cryptography, and so on.
The Burrows-Wheeler Transform (BWT) is a tool from Combinatorics on Words, that has found many applications well beyond its original purpose to be a preprocessing step for data compression.
In this talk, I present through examples some terms, notations, and concepts of Combinatorics on Words field. Words such as the Balanced words, Clustered words, Standard words possess special properties that can be used in order to understand more deeply the properties and the performances of the BWT. Finally, I show how these studies can be useful for BWT-based applications.
- 6 June 2017
Speaker: Omid Isfahani Alamdari
Title:Managing Spatio-Temporal Data of Moving Objects in Road Networks
Abstract: The proliferation of positioning devices such as smart phones, RFID tags and vehicle navigation systems and development in wireless technologies have resulted in an increasing growth in location-based services. Tracking of moving objects in different applications like traffic and transportation management systems, tourism and location-based social networks has resulted in massive amounts of data. The exponential increase in the amount of such trajectory data has caused communicational and storage problems and it is difficult to run spatio-temporal queries. This talk will cover a short introduction to an integrated method called “PCI” (Past Current Indexing) for indexing and storing spatial-temporal data of the past and present time modes simultaneously. This method uses map matching methods to remove noises and data reduction techniques to reduce storage and processing costs. In the last part of this lecture, a distributed indexing technique which is based on the MapReduce programming model will be presented. The efficiency of this index is experimented with different volumes of data and different number of computing nodes.
- 6 June 2017
Speaker: Luca Pedrelli
Title: Deep learning for sequences: Analysis of Hierarchical Linear Echo State Network
Abstract: After a recap of learning for temporal series within the
Reservoir Computing (RC) framework, we introduce a study on the role of layering in deep recurrent neural networks (RNNs). In this context, the use of deep Echo State Network (deepESN) with linear recurrent units allows us to bring more evidence on the intrinsic hierarchical temporal representation in deep RNNs through frequency analysis applied to the state signals. The potentiality of our approach is assessed on the class of Multiple Superimposed Oscillator tasks. Furthermore, our investigation provides useful insights to open a discussion on the main aspects that characterize the deep learning framework in the temporal domain.
- 13 June 2017
Speaker: Shima Moghtasedi
Title: Target Tracking in Wireless Sensor Networks: Kinetic Data Approach
Abstract. Tracking a target in wireless sensor networks is a well considered problem nowadays. In this problem the goal is to never loose the target while moving in a workspace covered by a wireless sensor network. Considering the energy consumption, the goal of reducing the energy used by the network becomes more noticeable and challenging. We study the problem of target tracking in wireless sensor networks with the aim of minimizing the number of used sensors. In this lecture I will present an event-based framework for target tracking problem and propose an online algorithm with optimal competitive ratio. Moreover, I observe the problem from the geometry perspective and define a coverage problem which is solved by an O(nlog n) time algorithm where n is the number of sensor’s neighbor to a query one. Finally to solve the coverage problem while objects (sensors) are moving with a known constant degree algebraic trajectory, a local and compact Kinetic Data Structure (KDS) is defined. I will show that the response time and efficient factor of the KDS for a query sensor is linear.
- 27 June 2017
Speaker: Roberto Trani
Title: From the query to the results: overview of the query processor pipeline
Abstract: Hundreds of millions of users each day look for the information they need on the Web using search engines trying to find out the relevant results from billions of indexed documents.
Moreover, the results must be returned in few tenths of second, because even a slight slow down may negatively impact the user engagement, resulting in fewer queries from the users and lastly in less revenues for the search engine.
Furthermore, queries submitted by users are short and intrinsically ambiguous since the terms that compose the queries can be inaccurate and general. Thus the relevant documents may not match the query. A common solution to deal with this issue is to use Query Expansion techniques that add terms to the original query to capture the user's original intent.
The aim of the seminar is to provide an overview of the entire query processor pipeline of a Web search engine: query expansion, retrieval and ranking.
- 4 July 2017
Speaker: Vinicius Monteiro
Title: Mining human mobility data and social media for smart mobility
Abstract: Mobility is a critical factor for quality of life in big cities. However, transport is also one of the most problematic aspects of urbanization.
In general, the unplanned urban growth of a city produces traffic congestion and challenges for transport services. Smart Mobility is a subset of smart initiatives, especially aiming at planning intelligent transport solutions, to achieve sustainable and satisfactory mobility strategies. Particularly focusing on the Ride-Sharing strategy, with the objective of reducing the number of circulating vehicles, we investigate a novel approach to boost ride sharing opportunities based on the knowledge of the human activity behind individual mobility demands.
- 11 July 2017
Speaker: Mariasole Bondioli
Title: Using technology to reduce anxiety in autistic children: problems and opportunities toward an innovative research challenge.
Abstract: During the last decade, several studies have introduced technology-enhanced systems for training autistic children or older subjects in communication, socialization, language and behavior. To better understand the problems and the opportunities of this kind of research, during the talk we will first introduce briefly the autistic spectrum disorder and the existent uses of technology to reduce distress in autistic children and adults. After that we summarize the challenges of some of the approaches exploiting technology present in the literature. Finally, we detail a case study in which we use technology to improve dental care in autistic patient. The project developed with the dentists of the S.Chiara hospital in Pisa has comes from a user study on 15 autistic children which will be summrized in the talk.