Risk management is the process of managing uncertainty through the identification, analysis, evaluation and treatment of risks. In a real scenario, risk management might have to deal with multiple stakeholders with different concerns and objectives. This means the concepts, techniques and practices suitable for its application might have a strong dependency from the alignment between the concepts, techniques and practices among the different context of the problem. Historically, in a first moment that lead to a proliferation of specialized frameworks defining how risk management should be established and implemented in specific domains, and therefore privileging viewpoints and views specific of those domains. However, if on one side those specialized solutions can prove to be effective, their specialization can also be a limitation to the practice of integrated risk management in scenarios of multiple heterogeneous domains, which is increasingly a common need in the enterprise world. Consequently, there is a need to solve the problem of integrating or simply reconciling several domains in a common context, a challenge recognized in the literature. In the system engineering discipline, the concepts of viewpoint an of view are used to represent the expression of a system from the concerns of a group of stakeholders. This way, one can reason that risk management can be a problem of alignment between different viewpoints and fragmented risk management views due to multiple stakeholders. This work addresses that problem by exploring the hypothesizes that a domain ontology can be used to both represent and manage those viewpoints, and therefore to support the integration and reuse of risk information. For that purpose, it is proposed a risk management domain ontology as also the principles for how to use that to define multiple viewpoints and views. The solution is demonstrated for a set of generic synthetic scenarios and validated for a real scenario of corporate risk management.

access_time 27 de maio de 2020 às 14:30
place https://videoconf-colibri.zoom.us/j/93407835298
local_offer Prova de Doutoramento
person Candidato: Ricardo João Correia Vieira
supervisor_account Orientador: Prof. José Luís Brinquete Borbinha

Digital technologies have been and continue to be incorporated into our day-to-day life after the popularization of the internet. They are introducing contemporary habits, novel ways of living together, new cultures, further shapes for the traditional businesses, and new kinds of businesses. To be successful in the Digital Age, the mere incorporation of Information Technology (IT) into business processes is not enough. Companies must be particularly agile and must reinvent themselves in three areas: business models, business processes, and work. A common understanding of this theme is a challenge for organizations. The problem that this thesis aims to address is how to support the organizations in aligning their digital business processes in Digital Transformation and Innovation (DT&I) initiatives, considering to what extent the business process management (BPM) is affected by, and affects digital transformation and innovation initiatives. This work considers, to solve the problem, having a model to guide and agile the development of the DT&I initiative, as well as minimizing its risk of failure. Thus, this thesis proposes a digital transformation and innovation business process management framework, which is composed of an ontology, a method, and a set of requirements for tools that automate this scope. The Design Science Research Methodology (DSRM) was chosen as support to solve this problem. Up to now, the following artifacts have been designed as contributions to understanding the problem: a quasi-systematic literature review, the ontology, a model aligned to the ArchiMate, and a new ArchiMate viewpoints proposal. The quasi-systematic literature review is used to identify, analyze, and synthesize the various aspects of the main concepts related to DI&T and the works related to this theme until then. The ontology is developed to provide a formal and explicit specification as a common understanding basis to support modeling DT&I initiatives. A Reference Model is provided to extend the ArchiMate with the main concepts identified in the ontology and to discuss to what extent ArchiMate would support the implementation of DT&I initiatives. Furthermore, a new ArchiMate viewpoint is proposed to extend and improve the reference model and propose the views to support its use. Future works address the following subjects: applying the Unified Foundational Ontology (UFO) as a foundation for the proposed ontology; developing a method to support the usage of the proposed ontology; specifying the functional requirements of a tool for supporting the method and the ontology.

access_time 20 de maio de 2020 às 15:00
place https://videoconf-colibri.zoom.us/j/97484790554
local_offer Prova de CAT
person Candidato: Silvia Bogéa Gomes
supervisor_account Orientador: Prof. Miguel Leitão Bignolas Mira da Silva / Dr.ª Flávia Maria Santoro

I will start by giving a brief overview of my DeepSPIN ERC project (https://deep-spin.github.io), whose goal is to develop new deep learning methods, models, and algorithms for structured prediction in natural language processing (NLP). Then, I will cover in more detail some recent work done in my group on sparse attention mechanisms. Attention mechanisms have become ubiquitous in NLP. Recent architectures, notably the Transformer, learn powerful context-aware word representations through layered, multi-headed attention. The multiple heads learn diverse types of word relationships. However, with standard softmax attention, all attention heads are dense, assigning a non-zero weight to all context words. In this talk, I will introduce the adaptively sparse Transformer, wherein attention heads have flexible, context-dependent sparsity patterns. This sparsity is accomplished by replacing softmax with alpha-entmax: a differentiable generalization of softmax that allows low-scoring words to receive precisely zero weight. Moreover, we derive a method to automatically learn the alpha parameter—which controls the shape and sparsity of alpha-entmax—allowing attention heads to choose between focused or spread-out behavior. Our adaptively sparse Transformer improves interpretability and head diversity when compared to softmax Transformers on machine translation datasets. Findings of the quantitative and qualitative analysis of our approach include that heads in different layers learn different sparsity preferences and tend to be more diverse in their attention distributions than softmax Transformers. Furthermore, at no cost in accuracy, sparsity in attention heads helps to uncover different head specializations. Joint work with Ben Peters, Gonçalo Correia, Vlad Niculae, Chaitanya Malaviya, Pedro Ferreira, Julia Kreutzer, Mathieu Blondel, Claire Cardie, Ramon Astudillo.

access_time 22 de abril de 2020 às 13:30
place Videoconference
local_offer DEI às quartas
face André Martins

We study the problem of multi-dimensional revenue maximization when selling $m$ items to a buyer that has additive valuations for them, drawn from a (possibly correlated) prior distribution. Unlike traditional Bayesian auction design, we assume that the seller has a very restricted knowledge of this prior: they only know the mean $\mu_j$ and an upper bound $\sigma_j$ on the standard deviation of each item's marginal distribution. Our goal is to design mechanisms that achieve good revenue against an ideal optimal auction that has full knowledge of the distribution in advance. We show that selling the items via separate price lotteries achieves an $O(\log r)$ approximation ratio where $r=\max_j(\sigma_j/\mu_j)$ is the maximum coefficient of variation across the items. If forced to restrict ourselves to deterministic mechanisms, this guarantee degrades to $O(r^2)$. Assuming independence of the item valuations, these ratios can be further improved by pricing the full bundle. We demonstrate the optimality of the above mechanisms by providing matching lower bounds. Our tight analysis for the deterministic case resolves an open gap from the work of Azar and Micali [ITCS'13]. As a by-product, we also show how one can directly use our upper bounds to improve and extend previous results related to the parametric auctions of Azar et al. [SODA'13]. This talk is based on joint work with Yiannis Giannakopoulos and Alexandros Tsigonias-Dimitriadis.

access_time 01 de abril de 2020 às 13:00
place Videoconference
local_offer DEI às quartas
face Diogo Poças

This thesis aims to improve remote collaboration in shared 3D workspaces. Current mixed reality technologies allow geographically distant collaborators to be together and share the samevirtual space, making it possible for people to see each other through realistic virtual representations. Face-to-face telepresence also promotes a sense of presence and can improve collaboration by allowing immediate understanding of nonverbal cues. Indeed, several approaches have successfully explored face-to-face remote interactions with 2D content. However, when collaborating in a 3D object-centered volumetric workspace, there is a decrease in awareness due to gesture ambiguities, occlusions, and different participants’ viewpoints. In this dissertation, we contribute the use of perception manipulation to improve workspace awareness in computer-supported collaborative work in mixed reality telepresence environments by assuring that remote collaborators are always aware of what is happening in the workspace when communicating using nonverbal cues. We began by contributing the technological foundations to prototype remote interactions. And then, we proposed and evaluated perception manipulation techniques focused on allowing remote people always to share the same understanding of the workspace. And, at the same time, being aware of nonverbal communication. Results suggest that by purposefully changing the properties of the person-task space using geometric transformations, warping, and repositioning devices, we can counteract gesture ambiguities, eliminate workspace occlusions, and promote a shared understanding of the workspace. In conclusion, we have validated our thesis, stating that perception manipulation techniques increase workspace awareness and improve face-to-face remote collaboration in mixed reality 3D workspaces.

access_time 31 de março de 2020 às 14:00
place https://videoconf-colibri.zoom.us/j/199216208
local_offer Prova de Doutoramento
person Candidato: António Maurício Lança Tavares de Sousa
supervisor_account Orientador: Prof. Joaquim Armando Pires Jorge