|
|
 |
|
Lundi 6 Mars
| Heure: |
14:00 - 15:00 |
| Lieu: |
Salle B107, bâtiment B, Université de Villetaneuse |
| Résumé: |
Automatic Deception Detection in Text Applying Topic Modeling Algorithms |
| Description: |
Hiram Calvo We deal with deceptive text identification by using different kinds of features: a continuous semantic space model based on latent Dirichlet allocation topics (LDA), one-hot representation (OHR), syntactic information from syntactic n-grams (SN), and lexicon-based features using the linguistic inquiry and word count dictionary (LIWC). We will present experiments with several combinations of these features were tested to assess the best source(s) for deceptive text identification aiming to present a state of the art performance. We conducted our tests on three different available corpora: a corpus consisting of 800 reviews about hotels, a corpus consisting of 600 reviews about controversial topics, and a corpus consisting of 236 book reviews. Additionally, we present an analysis on which features lead to either deceptive or truthful texts, finding that certain words can play different roles (sometimes even opposing ones) depending on the task being evaluated. We will present results of experiments in one-domain setting by training and testing our models separately on each dataset (with fivefold cross-validation); in a mixed-domain setting by merging all datasets into one large corpus (again, with fivefold cross-validation), and finally, with cross-domain setting: using one dataset for testing and a concatenation of all other datasets for training. |
Mardi 7 Mars
| Heure: |
11:30 - 12:30 |
| Lieu: |
Salle B107, bâtiment B, Université de Villetaneuse |
| Résumé: |
Valid quadratic inequalities for convex and some non-convex quadratic sets |
| Description: |
Julio César Góez In recent years, the generalization of Balas disjunctive cuts for mixed integer linear optimization problems to mixed integer non-linear optimization problems has received significant attention. Among these studies, mixed integer second order cone optimization (MISOCO) is a special case. For MISOCO one has the disjuncti ve conic cuts approach. That generalization introduced the concept of disjunctive conic cuts (DCCs) and disjunctive cylindrical cuts (DCyCs). Specifically, it showed that under some mild assumptions the intersection of those DCCs and DCyCs with a closed convex set, given as the intersection of a second order cone and an affine set, is the convex hull of the intersection of the same set with a linear disjunction. The key element in that analysis is the use of pencils of quadrics to find close forms for deriving the DCCs and DCyCs. In this talk we present an overview of the DCCs main results and we use the same approach to show the existence of valid conic inequalities for hyperboloids and non-convex quadratic cones when the disjunction is defined by parallel hyperplanes. Joint work with Miguel F. Anjos. |
Jeudi 9 Mars
| Heure: |
10:30 - 11:30 |
| Lieu: |
Salle B107, bâtiment B, Université de Villetaneuse |
| Résumé: |
Hybride Modelling, Analysis and Quantitative Verification of Large Biological Regulatory Networks |
| Description: |
Louis Fippo Fitime Biological Regulatory Networks (BRNs) are usually used in systems biology for modelling,
understanding and controlling the dynamics of different biological functions (differentiation,
proliferation, proteins synthesis, apoptose) inside cells. Those networks are enhanced with
experimental data that are nowadays more available which give an idea on the dynamics of BRNs components.
Formal analysis of such models fails in front of the combinatorial explosion of generated behaviours
despite the fact that BRNs provide abstract representation of biological systems.
This thesis handles hybrid modelling, the simulation, the formal verification and control of Large
Biological Regulatory Networks. This modelling is done thanks to stochastic automata networks, thereafter
to Process Hitting by integrating time-series data.
Firstly, this thesis proposes a refining of the dynamics by estimation of stochastic and temporal (delay)
parameters from time-series data and integration of those parameters in automata networks models. This
integration allows the parametrisation of the transitions between the states of the system. Then,
a statistical analysis of the traces of the stochastic simulation is proposed to compare the dynamics
of simulations with the experimental data.
Secondly, this thesis develops static analysis by abstract interpretation in the automata networks
allowing efficient under- and over-approximation of quantitative (probability and delay) reachability properties.
This analysis enables to highlight the critical components to satisfy these properties.
Finally, taking advantage from the previous developed static analyses for the reachability properties in the
qualitative point of view, and from the power of logic programming (Answer Set Programming), this thesis addresses the domain
of control of system by proposing the identification of bifurcation transitions. Bifurcations are
transitions after which the system can no longer reach a state that was previously reachable. |
Vendredi 10 Mars
| Heure: |
11:00 - 12:30 |
| Lieu: |
Salle B107, bâtiment B, Université de Villetaneuse |
| Résumé: |
Retrofitting linear types |
| Description: |
Arnaud Spiwack Type systems based on linear logic and their friends have proved that they are both capable of expressing a wealth of interesting abstractions. Among these the ability to mix garbage-collected and explicitly managed data in the same language. This is of prime interest for distributed computations that need to reduce latency induced by GC pauses: a smaller GC heap is a happier GC heap.
I had always had the belief that to add linear types to a language was a rather intrusive procedure and that a language with linear types would basically be an entirely new language. The Rust language is a case in point: it has a linear-like type system, but it's a very different language from your run-of-the-mill functional language.
This turns out not to be the case: this talk presents a way to extend a functional programming language. We are targeting Haskell but there is little specific to Haskell in this presentation. I will focus mostly on the type system and how it can be generalised: among other things the type system extends to dependent types. |
|
|
