| Links' Seminars and Public Events |   |
| 2021 | |
|---|---|
Fri 19th Feb 10:00 am 11:00 am | Seminar: Bernardo Subercaseau Title: Foundations of Languages for Interpretability. Abstract: The area of interpretability in Machine Learning aims for the design of algorithms that we humans can understand and trust. One of the fundamental questions of interpretability is: given a classifier M, and an input vector x, why did M classify x as M(x)? In order to approximate an answer to this "why" question, many concrete queries, metrics and scores have emerged as proxies, and their complexity has been studied over different classes of models. Many of these analyses are ad-hoc, but they tend to agree on the fact that these queries and scores are hard to compute over Neural Networks, but easy to compute over Decision Trees. It is thus natural to think of a more general approach, like a query language in which users could write an arbitrary number of different queries, and that would allow for a generalized study of the complexity of interpreting different ML models. Our work proposes foundations for such a language, tying to First Order Logic, as a way to have a clear understanding of its expressiveness and complexity. We manage to define a minimalistic structure over FO that allows expressing many natural interpretability queries over models, and we show that evaluating such queries can be done efficiently for Decision Trees, in data-complexity. Zoom link: univ-lille-fr.zoom.us/j/95419000064 |
Fri 12th Feb 10:00 am 12:00 pm | Seminar: Florent Capelli Title: Regularizing the delay of enumeration algorithms Zoom link: univ-lille-fr.zoom.us/j/95419000064 Abstract: Enumeration algorithms are algorithms whose goal is to output the set of all solutions to a given problem. There exists different measures for the quality of such algorithm, whose relevance depends on what the user wants to do with the solutions set. If the goal of the user is to explore some solutions or to transform the solutions as they are outputted with a stream-like algorithm, a relevant measure of the complexity of an enumeration algorithm is the delay between the output of two distinct solutions. Following this line of thoughts, significant efforts have been made by the community to design polynomial delay algorithms, that is, algorithms whose delay between the output of two new solutions is polynomial in the size of the input. While this measure is interesting, it is not always completely necessary to have a bound on the delay and it is enough to ask for a guarantee that running the algorithm for O(t poly(n)) will result in the output of at least t solutions. Of course, by storing each solution seen and outputting them regularly, one can simulate a polynomial delay but if the number of solutions is large, it may result in a blow up in the space used by the enumerator. In this talk, we will present a new technique that allow to transform such algorithm into polynomial delay algorithm using polynomial space. This is joint work with Yann Strozecki. |
Fri 15th Jan 10:00 am 12:00 pm | Séminaire de Kim Nguyễn Titile: The BOLDR project Abstract: I n this presentation, I will give an account of the BOLDR project and perspectives in the field of language integrated queries. Several classes of solutions allow programming languages to express queries: specific APIs such as JDBC, Object-Relational Mappings (ORMs) such as Hibernate, and language-integrated query frameworks such as Microsoft's LINQ. However, most of these solutions do not allow for efficient cross-databases queries, and none allow the use of complex application logic from the programming language in queries. We study the design of a new language-integrated query framework called BOLDR that allows the evaluation in databases of queries written in general-purpose programming languages containing application logic, and targeting several databases following different data models. In this framework, application queries are translated to an intermediate representation. Then, they are typed with a type system extensible by databases in order to detect which database language each subexpression should be translated to. This type system also allows us to detect a class of errors before execution. Next, they are rewritten in order to avoid query avalanches and make the most out of database optimizations. Finally, queries are sent for evaluation to the corresponding databases and the results are converted back to the application. Our experiments show that the techniques we implemented are applicable to real-world database applications, successfully handling a variety of language-integrated queries with good performances. This talk will give an overview of what has been achieved so far (mainly in the context of Julien Lopez' PhD Thesis) and will glimpse at preliminary work that is being done in the context of a collaboration with Oracle Labs. |
Fri 8th Jan 10:45 am 12:30 pm | Séminaire @ Lê Thành Dũng (Tito) Nguyễn Title: The planar geometry of first-order string transductions (joint work with Pierre Pradic) Abstract: hal.archives-ouvertes......ument We propose a new machine model recognizing star-free languages, with a geometric flavor. Our starting point is the characterization of regular languages using two-way automata (2DFA). The idea is to take seriously the visual representations found throughout the literature of the behavior of a 2DFA on a word ; by putting a total order on the set of states, one can formally define what it means for such a behavior to be planar, in a sense analogous to the planarity of combinatorial maps. Star-free languages are then exactly the languages recognized by "planar 2DFA". We also show that the corresponding planar transducer model characterizes the class of first-order transductions (a.k.a. aperiodic regular functions). If time allows, the talk will briefly discuss the connections of this work with the non-commutative lambda-calculus (cf. our recent paper Aperiodicity in a non-commutative logic, ICALP'20). |
