Monday 10 April 2023 |
||||||
|---|---|---|---|---|---|---|
| 10:00- 10:30 |
Open ceremony |
|||||
| 10:30-11:15 |
Fabio Cozman (USP) tbc Auditório Jacy Monteiro – IME –USP |
|||||
| 11:15-11:30 |
Break |
|||||
| 11:30-12:15 |
Frédéric Precioso
Université Côte d’Azur, Laboratoire I3S - UMR CNRS 7271, Inria Maasai team
|
|||||
| 12:15- 13:30 |
Lunch |
|||||
| 13:30-14:15 |
Roberto Kraenkel (UNESP)
|
|||||
|
Spotlight possibilities section
|
Mathematical Epidemiology section
|
|||||
| 14:15-14:45 |
Zakaria Benomar, Inria MiMove team (Paris) |
Pierre-Alexandre Bliman (Inria) |
||||
| 15:00-15:30 |
Fabio Kon (USP) |
Bedreddine Ainseba (Université de Bordeaux) |
||||
| 15:30-16:00 |
Patrick Valduriez
|
Max Souza (UFF) |
||||
| 16:00-16:45 |
Coffee break |
|||||
| 16:45-17:45 |
Visit to INOVA |
|
||||
| 17:45-18:15 | Visit to CCSL | Claudia Pio (UNESP) | ||||
| 18:15-18:45 |
Centro HPC em SP |
Helenice O F Silva (Unesp) |
||||
Tuesday 11 April 2023 |
|
|---|---|
|
Auditório Jacy Monteiro |
|
| 09:30- 09:45 |
Welcome and networking’s coffee |
| 09:45- 10:30 |
André C.P.L.F de Carvalho (ICMC- USP) |
| 10:30-11:15 |
Cooperation Programs: |
| 11:15-12:00 |
Future Collaboration: |
| 12:00-14:00 |
Almoço |
| 14:00-15:00 |
Round table science and innovation: |
Title: Research at the Center for Artificial Intelligence (C4AI)
Abstract: This talk describes activities at the C4AI, in particular as related to research. The center, sponsored by IBM and FAPESP, is committed to state-of-art research in Artificial Intelligence (AI), exploring both foundational issues and applied research. The C4AI also supports studies on the social and economic impact of AI and carries activities aimed at technology transfer and knowledge diffusion, looking for ways to improve human well-being and to increase diversity and inclusion.
Title: Covid-19 epidemic: models, and what to do with them.
In this talk I will introduce models used to quantitatively assess the Covid-19 epidemic, using examples and data from Brazil. I will discuss the role of the effective reproduction number, how to calculate it and in which situations it is useful as a policy instrument. I will also present how models where used to estimate the reproduction number and the number os re-infections during the Covid-19 gamma-variant outbreak in Manaus, Brazil, in January 2021. I will end with some considerations about the lessons learned during the pandemic and future prospects.
Title: Heuristic techniques applied to problems in Optimal Control
Abstract:This presentation gives an overview of heuristic optimization techniques, such as evolutionary algorithms (EAs), used to determine optimal control strategies. These heuristic techniques can solve very complex nonlinear optimization problems, which present computational difficulties or even cannot be handled by any analytic approaches. The advantage of metaheuristics over the classical optimization methods, such as Pontryagin's maximum principle (PMP), is the ease of adding constraints on the system and also the ability to work with different controls. These methods and some applications will be presented in this lecture.
Title: Forecasting the effect of Pre-Exposure Prophylaxis (PrEP) on HIV propagation with a system of delay differential equations
Abstract: The HIV/AIDS epidemic is still active worldwide with no existing definitive cure. Based on the WHO recommendations stated in 2014, a treatment, called Pre-Exposure Prophylaxis (PrEP), has been used in the world, and more particularly in France since 2016, to prevent HIV infections. In this work, we propose a new compartmental epidemiological model with a limited protection time offered by this new treatment. We describe the PrEP compartment with an age-structure hyperbolic equation and introduce a differential equation on the parameter that governs the PrEP starting process. This leads us to a nonlinear system with discrete delay. After a local stability analysis, we prove the global behavior of the system. Finally, we illustrate the solutions with numerical simulations based on the data of the French Men who have Sex with Men (MSM) population. We show that the choice of a logistic time dynamics combined with our Hill-function-like model leads to a perfect data fit. These results enable us to forecast the evolution of the HIV epidemics in France if the populations keep using PrEP. This work was done in collaboration with members of the Inria Dracula team, Laurent Pujo-Menjouet, Julien Molina, Grégoire Ranson and Jianhong Wu from York University.
Title: Modeling pest dynamics
Abstract:Pests develop strategies to maximize their survival. They can have one or more than one generation per year. Some individuals at the cocon stage can emerge one or two years after individuals of the same generation. In this talk we will show how one can modelize this problem. Then we will discuss the qualitative properties of the solutions and give some numerical simulations corresponding to different scenariis.
Title: A middleware-based approach for enabling interoperable IoT-to-Edge-to-Cloud environments
Abstract:The explosive growth and increasing power of IoT devices, along with the rise of advanced communication options (e.g., 5G), have resulted in unprecedented volumes of data. To manage and process this massive amount of data, such tasks are often delegated to centralized computing facilities (i.e., Cloud platforms). Albeit providing an abundance of resources, the centralized nature of the cloud can induce a significant topological distance between the computing resources hosted on the Cloud and the vast majority of IoT devices due to their geographical location. Thus, one of the biggest challenges that the Cloud is facing is to address the increasing demands for new services in terms of Quality of Service (QoS), such as computational speed and faster response times. To deal with this drawback, multi-layered architectures are emerging, where computing resources and applications are distributed from the edge of the network to the cloud, realizing the IoT-to-Edge-to-Cloud computing continuum. In this architecture, the fact that the Edge computing nodes are topologically in proximity to end devices is a key enabler of advanced services/applications that were not feasible before while relying solely on the Cloud. Data can be processed at the Edge and then offloaded to the Cloud, when required, for further processing. Yet, deploying applications using the continuum's distributed infrastructure is very challenging due to the geo-distribution of the infrastructure (i.e., networking issues), the heterogeneity of underlying hardware/technologies, and the lack of efficient abstractions for handling and orchestrating diverse workloads. This presentation will focus on the work being conducted by the Middleware on the Move (MiMove) team at INRIA, Paris, to address the interoperability issues in an IoT-to-Edge-to-Cloud Continuum environment.
Title: Minimizing epidemic final size through social distancing
Abstract: How to apply partial or total containment measures during a given finite time interval, in order to minimize the final size of an epidemic — that is the cumulative number of cases infected during its whole course? We provide here a complete answer to this question for the SIR epidemic model. Existence and uniqueness of an optimal strategy is proved for the infinite-horizon problem corresponding to control on an interval $[0,T]$, $T>0$ (Problem 1), and then on any interval of length $T$ (Problem 2). For both problems, the best policy consists in applying the maximal allowed social distancing effort until the end of the interval $[0,T]$ (Problem 1), or during a whole interval of length $T$ (Problem 2), starting at a date that is not systematically the closest date, and which may be computed by a simple algorithm. These optimal interventions have to begin before the proportion of susceptible individuals crosses the herd immunity level, and lead to limiting values smaller than this threshold. We also study the following more general issue (Problem 3): how to apply confinement during a (possibly disconnected) set of measure at most $T$, in order to minimize the final size? This problem is shown to have the same optimal solution than Problem 2. These results have been obtained with Michel Duprez (Inria), Yannick Privat (Université de Strasbourg), Alain Rapaport (INRAE) and Nicolas Vauchelet (Université Sorbonne Paris Nord).
Title: Many ways of combining Symbolic and Non-symbolic AI
Abstract: In these works, we explore several ways to combine Symbolic and Non-symbolic AI. We are convinced, as several other colleagues in AI community, that reuniting these two branches of IA would lead to impressive, more robust, more reliable, and more sustainable, AI models. I will present some recent works we have done on Kownledge-Driven Active Learning, where the criterion to select which new unlabeled sample should be annotated directly derives from prior external knowledge on the data and First-Order-Logic formulation. I will also present our on-going work on "Learning to Explain and Explaining to Learn" where we look for concepts learned by the neural network and for designing new hierarchical losses taking into account any priori knowledge. I will finally present our recently published work on Concept Model Embedding, and explain why the results are important and what could be the next step. Finally, I will conclude with some applications and contexts of application, in particular in automotive.
Title: Data and Model Management with Gypscie*
Abstract: To realize the full potential of data science, models (in particular, ML models) must be built, combined and ensembled, which can be very complex as there can be many models to select from. Furthermore, they should be shared and reused, in particular, in different execution environments such as HPC or Spark clusters. To address this problem, we propose Gypscie, a new framework that supports the entire ML lifecycle and enables model reuse and import from other frameworks. The approach behind Gypscie is to combine several rich capabilities for model and data management, and model execution, which are typically provided by different tools, in a unique framework. *Joint work with Fabio Porto, LNCC, Brazil
Title: Future Internet, Software Systems, and Smart Cities
Abstract: The National Institute of S&T of the Future Internet for Smart Cities (also known as the InterSCity project) is hosted at IME-USP and carries out research on Distributed Internet Systems and the construction of Complex Software Systems; many of the applications target urban problems, in what is known as Smart Cities. In this talk, we will discuss the major scientific and technological challenges that are being tackled by InterSCity researchers as well as some of the results achieved by them in recent years. Finally, we will discuss future research directions in the field.
Title: Modelling communicable diseases linked to real data
Abstract: The inter-host modelling of communicable diseases is a very active research area that have gained wide visibility during COVID-19 pandemics. The dynamical models are as much complex as the relevant ingredients introduced in order to catch the features associated with the research question. Even more the validation of a constructed model is related to to its capacity of describing real epidemic and endemic data as well as of designing different scenarios to investigate hypothetical situations, such as vaccination strategies [1]. In this talk I will present some inter-host models that we have developed for directly transmitted diseases [2] and vector-borne transmitted diseases [3] linked to real data of epidemics in Brazilian urban centers. I also explore some heterogeneous scenarios, for instance, of co-circulating viruses, such as in simultaneous Dengue and Zika epidemics that had occured in Brazil in 2015-2016. Finally I will also show our recent methodological result [4] for estimating effective reproduction number R(t) that consist in a generalization of the well-known next generation method used to calculate the basic reproduction number R0 of different epidemiological models.
Title: Exploring the impact of temperature on the efficacy of replacing the wild Aedes aegypti population by Wolbachia-carrying one
Abstract: In this talk, we will present and discuss the results obtained by a non-autonomous time-delayed differential system, which was proposed to reproduce the dynamics of both Wolbachia-infected and non-infected mosquito populations in several scenarios that differ by daily environment temperature, the bacterial strain that promotes the infection on mosquitoes, and the release guidelines of infected mosquitoes. The numerical results show that: (i) multiple releases were more efficient than a single one, (ii) when the mosquito population is high is the best moment to implement the releasing of infected mosquitoes, (iii) strains that produce both high levels of cytoplasmic incompatibility and maternal inheritance boost the efficacy of the technique, (iv) high temperature can jeopardize the efficacy of the technique.
Title: IARA - Brazilian Applied Artificial Intelligence Research Center for Smart Environments
Abstract: As results of the Brazilian Artificial Intelligence Strategy, 8 National Research Centers on Applied Artificial Intelligence are being created, each center dedicated to at least 1 of 5 areas: Agriculture, Cybersecurity, Industry, Health and Smart Cities. 6 of these centers were announced in 2021, 2 for industry, 3 for health and 1 for smart cities, named IARA. Just as economic and social advances allow people to live longer and better, they also raise awareness that these advances need to happen faster and to be more sustainable and inclusive. This generates a growing demand for a better quality of life, both in urban and rural regions. One of the main paths for this increase is through the incorporation of new technologies, which allow, for example, the transformation of cities into smart cities. IARA, Artificial Intelligence for Smart Environments has almost 200 PhD researchers from more than 40 Brazilian Universities and Research Centers, support from 21 internacional universities/research centers, and 7 companies. With the inicial support from 8 Brazilian cities, IARA started with 2 cities, Canaã dos Carajás, north of Brazil, and Guarapuava, south of Brazil.
Institute of Mathematics and Statistics
University of São Paulo
Rua do Matão, 1010, Cidade Universitária, São Paulo, Brazil
ibw2023@ime.usp.br