BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Date iCal//NONSGML kigkonsult.se iCalcreator 2.20.2//
METHOD:PUBLISH
X-WR-CALNAME;VALUE=TEXT:Eventi DIAG
BEGIN:VTIMEZONE
TZID:Europe/Paris
BEGIN:STANDARD
DTSTART:20231029T030000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20230326T020000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:calendar.26189.field_data.0@www.u-gov-ricerca.uniroma1.it
DTSTAMP:20260404T184654Z
CREATED:20230708T202141Z
DESCRIPTION:Linear Temporal Logic (LTL) is a modal logic widely used in dif
 ferent domains\, such as robotics and Business Process Management\, for sp
 ecifying temporal relationships\, dynamic constraints\, and performing aut
 omated reasoning. However\, exploiting LTL knowledge in real-world applica
 tions can be difficult due to the knowledge's symbolic 'crispy' nature. Th
 is seminar explores different techniques to relax the knowledge to make it
  applicable in continuous domains where symbols are grounded through Deep 
 Learning modules and the symbol grounding function and/or the symbolic tem
 poral specification can be unknown or partially known. In particular\, we 
 propose two different techniques: (i) one based on Logic Tensor Networks a
 nd (ii) one based on Probabilistic Finite Automaton. We apply the first ap
 proach to classifying sequences of images\, and we show that our approach 
 requires less data and is less prone to overfitting than purely deep-learn
 ing-based methods. We use the second approach to learn DFA specifications 
 from traces with gradient-based optimization\, showing that it can learn l
 arger automata and is more resilient to noise in the dataset than prior wo
 rk. Finally\, we propose an extension of our second approach that we apply
  to non-Markovian Deep Reinforcement Learning problems. This third contrib
 ution has shown to be more sample efficient of methods based on Recurrent 
 Neural Networks\, and\, at the same time\, it requires less prior knowledg
 e than methods based on LTL\, such as Reward Machines and Restraining Bolt
 s.
DTSTART;TZID=Europe/Paris:20230710T110000
DTEND;TZID=Europe/Paris:20230710T110000
LAST-MODIFIED:20230709T174702Z
LOCATION:Aula Magna
SUMMARY:Seminario pubblico di Elena Umili (Procedura valutativa per n.4 pos
 ti di Ricercatore a tempo determinato tipologia A - SC 09/H1 SSD ING-INF/0
 5) - Integrating Linear Temporal Logic with Deep-Learning-Based Applicatio
 ns - Elena Umili
URL;TYPE=URI:http://www.u-gov-ricerca.uniroma1.it/node/26189
END:VEVENT
END:VCALENDAR