Homepage‎ > ‎


Research Lines (suggested/open subjects highlighted)
  • Cooperative Active Perception
  • Using Agent Reinforcement in Open Neuroevolutionary Controler Scenarios (co-supervision Prof. Sancho OliveiraBioMachines Lab, starting Jun 2017, finish Jul 2018)
  • Intelligent Home Automation (colaborations with Muzzley)
    • Home usage patterns: Discover the patterns in the usage of a family home (MSc student João Cruz)
    • IoT buyer patterns analysis (co-supervision w/ Prof. Paulo Rita) (starting Jun - Sep 2017, finish Jul 2018)
    • Dynamic App: usage patterns and user-adaption (co-supervision w/ Prof. Isabel Alexandre) (starting Jun - Sep 2018, finish Jul 2019)
  • Urban Sensing (Data Fusion and Urban Sensor Data analysis and prediction)
    • Intelligent Transport Systems / Decision Support for Public Space Management 
      • Parking spaces prediction (data courtesy of EMEL, MSc student Marco Silva)
      • Detection of motorcycle driver stress (MSc student Fábio Dias)
    • Analysis of elderly usage of public space (project under submission, possible scholarship, starting Jun - Sep 2018, finish Jul 2019)
    • Data fusion portal for real-estate information (MSc student João Andrade)
  • Predictive Analytics / Supervised Learning / Data Mining
    • Hotel Revenue Management: cancelation prediction, (data courtesy of STR, PhD student Nuno António)
    • CV screening, real data, cooperation with Siemens PT, co-supervision with Prof. Sérgio Moro  (MSc. student Mario Rivotti,  sholarship, starting Jun - Sep 2017, finish Jul 2018)
    • UsingText Mining for Stock Exchange variation prediction (co-supervision Prof. Ricardo Ribeiro) (MSc student José Serro)
  • Classification of multivariate data sequences applied to behavior analysis based on movement data (PhD student David Jardim, collaboration with Microsoft PT). Dados disponíveis em: https://github.com/DavidJardim/precog_dataset_16
    • Classification of customer behaviour based on multiple sensor data (project under submission with AXIANS, possible scholarship, starting Jun - Sep 2018, finish Jul 2019)
  • Automated structuring of learning tasks: sub-goal discovery
    • Open-source RL / HRL library (undergrad. André Freire)
    • Hierarchical and Multi-objective RL
  • Embodied cognition (generating artificial self-awareness)
Proposals TSI - 2017:
  • Colaboration on ongoinig work on the adaptation of a HRL (Hierarchical Reinforcement Learning) algorithm for a Reinforcement Learning open source platform (BURLAP, http://burlap.cs.brown.edu/). - João Gonçalves
  • Colaboration on ongoing research on Cooperative Active Perception using JBotEvolver, on evolving the following behaviours:
    • Formations on noisy and interruptive environments - Diogo Pires
    • Surrounding a prey (at a safe distance) - João Januário
  • Implementation for Weka open-source of version of the clustering algorithm defined in [Vicente & Nunes 2010]
Areas: Machine Learning, Data Mining, Data Science, Complex Systems (Learning agents in)

Other research interests 
  • Classification of multivariate data sequences

Some tips for starting an MSc / PhD:
  • presentation to see if you are planing to write a thesis in the near future.
  • Lookup "systematic literature review" and read about it.
  • A guide to writing your thesis, by António Dias Figueiredo: www.researchgate.net/publication/288670054
  • Tipical timeline for MSc: Oct - Dec: State of The Art (including finding and testing technical options); Jan - Feb: Build prototype; Mar - Jun: Experiments (repeat, repeat, repeat, ... ); Nov - Jul: Writing paper and thesis (yes ... writing starts in November! not in June, and yes, there should be a paper);
  • Conference calls: WikiCFP.
  • Starting a thesis: Write your motivation (why is it a problem), objectives (what is the problem you plan to solve), question (what don't you know now, that you plan to know after the thesis), approach (how do you solve the problem), expected results (how do you convince yourself and a jury that you solved the problem). Write what you know about the matter. Start researching more about it. Keep writing about what you find.
  • If enrolling on ISCTE-DCTI for PhD, read http://dcti.iscte.pt/doutoramentos/ pay special attention to the guide for writing a state-of-the-art (pt)
  • A few guidelines on the use of citations and references and a word about fraud in academic work (pt)
  • The graphical rules for writing a thesis in ISCTE, version approved in 2010 (at the bottom of this page, document named "Harmonizacao_Doc_ ... doc", in pt).
  • A class for making it easier to write a MSc thesis in the style required at ISCTE is also available below (ThesisTemplateLatex.zip) courtesy of Prof. André Santos, based on initial code by Prof. Miguel Duarte. Disclaimer: There are no guaranties of compliance to current standard.
  • Some of the best MSc/PhD thesis I have seen:
  • If you are planning to write in LaTeX try OverLeaf (https://www.overleaf.com/) or other collaborative LaTeX editors. At the bottom of this page you can find a class that resembles ISCTE'S pattern. This example was graciously made available by Prof. Carlos Serrão with no garanty whatsoever.
  • If I am supervising: Install Mendeley and, after my agreement to supervise, add me to your contacts and share your thesis-bibliography folder. Keep an eye out for news, I'll drop some files into that folder from time to time.
  • More guidelines, and some usefull tools (EasyBibJabRef, AnyStyle).
  • A humorous, but interesting view on graduated work (mostly for PhD candidates): (Azuma 1997-2003): Everything I wanted to know about C.S. graduate school at the beginning but didn't learn until later.

Involvement (currently) in funded projects:
  • none

Urban Sensing (M.Sc., colaborations with Muzzley)
  • Data-mining over sensor data for home and city sensors
  • Summary: Data Mining is the extraction of patterns and knowledgs from structred data. Usages range from determinig user habits, home security, determinig architetural obstacles, modeling complex user behaviors.
  • Relevant publications / conferences: PETS
  • Initial references:

Cooperative Active Perception (M.Sc., collaboration with BioMachines Lab)
  • 1 year, full-time,
  • Summary: Decentralized decision making is difficult to pre-program and, particularly, in some environments the only viable solution is to develop specific behaviors for particular problems.Even though a lot as been researched, the application of active perception in aquatic environments has specific problems that have not been addressed.
  • Applications: Swarm intelligence, Cooperative Active Perception 

Automated structuring of learning tasks: sub-goal discovery (M.Sc./Ph.D.) 
  • 1 to 3 years, full-time,
  • Summary: Goal discovery and task sub-division as an add-on to several Machine Learning algorithms is currently a major topic of research. Either as a boost to Reinforcement Learning, with Hierarchical Reinforcement Learning (HRL) or as way to improve the efficiency of Evolutionary Algorithms, new techniques to divide and conquer problems are surfacing. Yet, the automated discovery of subtasks (or selection of subgoals)? still has several interesting challenges, each of which can be directed to a different MSc (or explored more in depth by a PhD).
  • Important publications / conferences on the subject: NIPS, JMLR, IJCAI, AAMAS
  • Influential researchers / teams: Andrew Barto (UMass), Richard Sutton (U. Alberta)Bernard Hengst (UNSW)Peter Stone (UTexas)
  • Initial references:
    • Barto, A. G., & Mahadevan, S. (2003). Recent advances in hierarchical reinforcement learning. Discrete Event Dynamic Systems, 13(4), 341-379. Retrieved from http://www-anw.cs.umass.edu/pubs/2003/barto_m_DEDS03.pdf.
    • Sutton, R., Precup, D., & Singh, S. (1999). Between MDPs and semi-MDPs: A framework for temporal abstraction in reinforcement learning. Artificial Intelligence, 112, 181-211. Retrieved from http://webdocs.cs.ualberta.ca/~sutton/papers/SPS-aij.pdf.
    • David Jardim (M.Sc.) - "Hierarchical Reinforcement Learning", M.Sc. thesis, Instituto Universitário de Lisboa (ISCTE), 2010.
    • M. Duarte, S. Oliveira, and A. L. Christensen (2014), "Evolution of Hierarchical Controllers for Multirobot Systems", Proceedings of the Fourteenth International Conference on the Synthesis and Simulation of Living Systems (ALIFE), MIT Press, Boston, MA, pages 657-664

Behavior analysis based on movement data (Ph.D., collaboration with Microsoft PT).
  • 1 to 3 years, full-time,
  • Summary: Classify actions and behaviors (composed of several actions), predict future actions based on common structure of action-sequences, group similar actions or sequences. Data gathered using Microsoft's KINECT or other sources.
  • Relevant publications:
    • Jardim, D., Nunes, L., & Dias, M. (2015). Human Activity Recognition and Prediction. In Doctoral Consortium on Pattern Recognition Applications and Methods (DCPRAM 2015) (pp. 24–32). doi:10.5220/0005327200240032
  • Initial references:NA

Intelligent Transport Systems (M.Sc./Ph.D.)
Luís Nunes,
Jul 7, 2016, 12:59 AM
Luís Nunes,
Mar 21, 2012, 3:25 PM
Luís Nunes,
Sep 1, 2016, 8:37 AM
Software - Integrated.zip
Luís Nunes,
Jun 7, 2016, 4:04 AM
Luís Nunes,
Sep 7, 2016, 2:07 AM