진학프로석·박사 채용 접수 플랫폼

Scientific Advisor: Maria do Rosário De Oliveira Silva (ist12954)
Co-advisor(s): Maria do Rosário De Oliveira Silva (ist12954), CEMAT e Departamento de Matemática, Instituto Superior
Técnico, Universidade de Lisboa; Gonçalo Marques Fernandes de Oliveira (ist156809), CAMGSD e Departamento de
Matemática, Instituto Superior Técnico, Universidade de Lisboa; José Manuel Vergueiro Monteiro Cidade Mourão
(ist12816), CAMGSD e Departamento de Matemática, Instituto Superior Técnico, Universidade de Lisboa; Henrique
Manuel Dos Santos Silveira de Oliveira (ist13299), CAMGSD e Departamento de Matemática, Instituto Superior Técnico,
Universidade de Lisboa; Jorge Filipe Duarte Tiago (ist90590), CEMAT e Departamento de Matemática, Instituto Superior
Técnico, Universidade de Lisboa; Maria da Conceição Esperança Amado (ist13493), CEMAT e Departamento de
Matemática, Instituto Superior Técnico, Universidade de Lisboa; Erida Gjini (ist428854), CEMAT, Instituto Superior
Técnico, Universidade de Lisboa; Luís Carlos Costa Pinheiro de Carvalho (ist430352), CAMGSD, Instituto Superior Técnico,
Universidade de Lisboa e Departamento de Matemática (ISTA), ISCTE; Joao Lopes Costa (ist90642), CAMGSD, Instituto
Superior Técnico, Universidade de Lisboa e Departamento de Matemática (ISTA), ISCTE.
Organic Unit: Centre for Computational and Stochastic Mathematics
Scholarship Theme: A Forecasting Framework for Pandemic Monitoring and Decision-Making
Duration: 6 months
Maximum Duration Including Renewals: 6 months

Objectives
To develop and validate a hierarchical, time-dependent forecasting framework for the COVID-19 pandemic, integrating
heterogeneous epidemiological and data-driven models through a mixture-of-experts architecture that adaptively
combines specialized models according to pandemic phase and intervention regime. The framework aims to provide
robust, uncertainty-aware forecasts to support adaptive public health decision-making and resource planning. The work
is organized into four main tasks, aligned with the project's specific objectives:
(i) Data infrastructure. Build a reproducible pipeline integrating epidemiological, vaccination, intervention, and variant
data, with a temporal structure suitable for time-dependent modelling.
(ii) Portfolio of expert models. Develop and fit a diverse set of time-dependent forecasting models (e.g., neural network,
mechanistic, statistical, and data-driven) to serve as experts within the integrative architecture.
(iii) Mixture-of-experts integration. Design and implement a context-aware gating mechanism that adaptively combines
the expert models over time, addressing leakage-free training and uncertainty propagation.
(iv) Validation and benchmarking. Assess the framework with time-aware validation protocols and probabilistic error
metrics, benchmarking it against individual experts and standard ensembles across pandemic phases and scenarios.
Work Plan
The work is organized into four main tasks, aligned with the project's specific objectives:Task 1: Data infrastructure and literature review. Review the state of the art on time-dependent pandemicforecastingand mixture-of-experts architectures, and build a reproducible data pipeline integrating epidemiological,vaccination,intervention, and variant data from national and international sources.Task 2: Development of expert models. Develop and fit a diverse portfolio of time-dependent forecasting models(e.g.,neural networks, mechanistic, statistical, and data-driven) including models produced within the broader researchgroup,structured to serve as experts within the integrative architecture.Task 3: Mixture-of-experts integration. Design and implement a context-aware gating mechanism thatadaptivelycombines the expert models over time, addressing leakage-free training through rolling-origin protocols andthepropagation of predictive uncertainty.Task 4: Validation, benchmarking, and dissemination. Evaluate the framework using time-aware validation protocolsandprobabilistic error metrics, benchmark it against individual experts and standard ensembles across pandemic phasesandscenarios, and prepare the results for scientific dissemination and publication.

Contest Procedure
Applications must be exclusively submitted on the
admissions platform
of the
Instituto Superior Técnico
at
https://fenix.tecnico.ulisboa.pt/fenixedu-admissions 
and requires registration and validation of the candidate's identity.
Applications are only accepted when the form available in the platform is correctly filled, submitted and locked withoutany validation errors. The mandatory documentation to submit in the scholarship aplication includes:
Curriculum Vitae
Proof of Qualifications (or declaration of honor in case you do not yet have the certificate)
Proof of Registration/Enrolment
Motivation Letter
The application submission deadlines can be viewed in the admissions platform.
The results of the contest will be made available in the same admissions platform.