Português English
Contato
  • Série de Seminários do INF

    • Série de Seminários do INF | Seminários 2026/1

    24/06/2026

    The only Constant is Change

    • Local: Auditório 0
    • Título: The only Constant is Change
    • Palestrante(s): Prof. Marinho Barcellos
    • Resumo: In this talk, I will first share some reflections on life changes, from disruption to evolution, drawing on personal experiences, including an international move at the age of 55. I will discuss the associated trade-offs and challenge some common myths. In the second part of the talk, I will explore why, in cybersecurity, the only constant is change, and how this is reflected in some of my recent and ongoing research on Internet security.
    • Mini-bio: Marinho Barcellos is an Associate Professor at the University of Waikato, New Zealand, where he leads research on Internet security. Previously, he was a faculty member at UFRGS (2010–2019) and Unisinos (1998–2008), both in Brazil. He holds a PhD in Computer Science from the University of Newcastle upon Tyne, UK (1994–1998). Throughout his career, Marinho has disseminated research advances and contributed technical expertise to a range of scientific communities, including SBC SBRC/SBSeg, IEEE IM/NOMS/CNSM, IEEE ICC/GLOBECOM, and ACM SIGCOMM/CoNEXT/IMC/PAM.

    02/06/2026

    Distributed Model Training on Edge Devices under System and Data Heterogeneity

    • Local: Auditório 0 (palestra em inglês)
    • Título: Distributed Model Training on Edge Devices under System and Data Heterogeneity
    • Palestrante(s): Prof. Yasaman Amannejad, PhD
    • Resumo: In this talk, I will briefly introduce my broader research program in performance engineering and service management. Then, the main focus of the talk will then be on our recent work on Split Federated Learning under joint system and data heterogeneity. Split Federated Learning can be used to train machine learning models in a distributed way on edge devices. In these settings, devices may differ in computational capacity, network conditions, and local data characteristics, making uniform training configurations difficult to deploy efficiently. I will present the results of our study on Split Federated Learning with customized cut layers, where clients can use different cut layers based on their resource constraints while still contributing to a unified global model. This work is part of our ongoing research on resource-aware distributed learning systems, including optimal cut-layer selection based on device resources, model accuracy, and overall system performance.
    • Mini-bio: Yasaman Amannejad is an Associate Professor and Academic Director of the School of Computing Sciences and Mathematics at Mount Royal University in Calgary, Canada. Yasaman’s research is in performance engineering and service management. Her work includes contributions to workload generation and characterization, performance testing, execution-time prediction, and resource allocation. She has worked across a range of application domains, including web and cloud-based systems, big data processing systems, and, more recently, on machine learning applications. Her research has been supported by Canadian funding agencies and industry-facing programs, including NSERC, NFRF, and Mitacs. In addition to her research, Yasaman is active in research service and professional community building. She serves as a reviewer and chair for major Canadian research funding programs and contributes to IEEE activities in Canada as Chair of the IEEE Computer Society in Southern Alberta and as coordinator of Computer Society activities in Western Canada.