Le prochain séminaire aura lieu le 14 Février 2025

Les 3 derniers séminaires du LIM

Stable-driven stochastic models and their real-world applications

Animé par Solym Manou-Abi, le 06 Mars 2025 à 14:00:00

In this talk, I focus on stable driven stochastic models used to describe real-world phenomena, such as those in epidemiology and financial risk. The main objective is to address the estimation of the associated parameters and the stochastic approximation of these models. Références: [1] Hajjaji, O., Manou-Abi, S. M., & Slaoui, Y. (2024). Parameter estimation for stable distributions and their mixture. Journal of Applied Statistics (Taylor & Francis), 1–34. https://doi.org/10.1080/02664763.2024.2434627 [2] Solym M. Manou-Abi (2024). Parameter estimation for a class of stable driven stochastic differential equation. J Indian Soc Probab Stat (Springer). https://link.springer.com/article/10.1007/s41096-024-00196-7 [3] Solym M. Manou-Abi. Approximate solution for a class of stochastic differential equation driven by stable processes (2024) Under revision. https://hal.science/hal-04217398v3 [4] A. Joulin and S.M. Manou-Abi (2015). A note on convex ordering for stable stochastic integrals. Stochastics (Taylor & Francis), 87(4) : 1-12, 2015. https://www.tandfonline.com/doi/full/10.1080/17442508.2014.989528

Integrating Natural Language Processing with Research based Teaching and Learning in Developing Swahili Corpus

Animé par Abubakar Diwani Bakari, le 14 Février 2025 à 14:00:00

Natural language processing is a machine-learning technique that consists of multiple topics and tasks related to data and information management, leading to additional value in the teaching and learning process. Research-based teaching and learning is a constructive approach in which students learn through research, and instructors support the process by teaching in a way that does not present topics as closed but opens them up for students to research. This technique motivates students to learn collaboratively and communicate effectively. This study aimed to employ a new research-based teaching approach in the NLP domain for third-year students who opt for advanced topics in computer science courses to enhance their learning experience. Initially, the course only provided basic research foundations. Still, for the last two academic years (2021/2022 and 2022/2023), the module was dedicated to the current university effort to develop the Swahili Language corpus. In the new approach, teams of students are provided with topics, basic guidelines, and other aids to help work with different areas of research in Swahili Language structures. Student teams then deployed different libraries, tools, methods, and algorithms to solve big data Swahili text summarization, which is an ideal problem for addressing NLP learning. Incorporating these research-focused teaching and learning methods has empowered students, who have become critical thinkers, influential researchers, and lifelong learners. The outcome is a coherent and readable summary of research reports between 3000 and 5000 words based on the designed template and algorithms. It was observed that the adapted method suited the specific subject matter by setting the baseline that will feature the blueprint of the mega development of the Swahili language corpus. Abubakar Diwani Bakari Department of Computer Science & IT The State University of Zanzibar

Pure maps are strict monomorphisms

Animé par kristof kanalas, le 06 Février 2025 à 14:00:00

Categorical logic is 1. the algebraization of positive logic. The following are synonyms: h-inductive theory, coherent theory, ω-geometric theory. They can be encoded into small coherent categories ("fat distributive lattices"), and this way we can see theories as easy-to-handle mathematical objects. 2. a framework for doing abstract model theory. Those categories which arise as the category of models and homomorphisms of some theory in the geometric fragment of L∞,κ can be characterized. They are called accessible categories and they provide a setting where the toolkit of (infinitary, positive) model theory and that of category theory can be used simultaneously. I will try to illustrate how techniques coming from the syntactic side of categorical logic (number 1) can solve problems coming from the semantic side (number 2). In the book [2] the authors ask if it is true that in a λ-accessible category A, the λ-pure maps are strict monomorphisms. In [1] a counterexample is given. I will sketch the proof of two positive results: 1) if A is axiomatizable in finitary coherent logic, then the answer is affirmative and 2) if there is a proper class of strongly compact cardinals then for every λ-accessible category A there is some μ > λ, such that when we see A as a μ-accessible category then the answer is affirmative (that is: μ-pure maps are strict monomorphisms). The proof sketch will allow me to mention some powerful methods coming from category theory (motivated by geometry): A is sitting inside the λ-pro-completion of some small category, and to analyze this inclusion one has to understand Boolean-valued models from a sheaf theoretic perspective, following [4]. The talk is based on [3]. [1] J. Adamek, H. Hu, and W. Tholen. "On pure morphisms in accessible categories". In: Journal of Pure and Applied Algebra 107.1 (1996), pp. 1--8. [2] J. Adamek and J. Rosicky. Locally Presentable and Accessible Categories. Vol. 189. London Mathematical Society Lecture Note Series. London: Cambridge University Press, 1994. [3] Kristof Kanalas. Pure maps are strict monomorphisms. https://arxiv.org/abs/2407.13448. 2024. [4] Jacob Lurie. Lecture notes in Categorical Logic. 2018. url: https://www.math.ias.edu/~lurie/278x.html.