Conference Papers

Probabilistic object identification through on-demand partial views

Brandao S., Veloso M., Costeira J.P.

VISAPP 2014 - Proceedings of the 9th International Conference on Computer Vision Theory and Applications

2014

Abstract:

The current paper addresses the problem of object identification from multiple 3D partial views, collected from different view angles with the objective of disambiguating between similar objects. We assume a mobile robot equipped with a depth sensor that autonomously grasps an object from different positions, with no previous known pattern. The challenge is to efficiently combine the set of observations into a single classification. We approach the problem with a sequential importance resampling filter that allows to combine the sequence of observations and that, by its sampling nature, allows to handle the large number of possible partial views. In this context, we introduce innovations at the level of the partial view representation and at the formulation of the classification problem. We provide a qualitative comparison to support our representation and illustrate the identification process with a case study.

Products of stochastic matrices: Exact rate for convergence in probability for directed networks

Bajovic D., Xavier J., Sinopoli B.

2012 20th Telecommunications Forum, TELFOR 2012 - Proceedings

2012

Abstract:

We study the products W k ···W 1 of random stochastic, not necessarily symmetric matrices. It is known that, under certain conditions, the product W k · · · W 1 converges almost surely (a.s.) to a random rank-one matrix; the latter is equivalent to |λ 2 (W k · · · W 1 )| → 0 a.s., where λ 2 (·) is the second largest (in modulus) eigenvalue. In this paper, we show that the probability that |λ 2 (W k · · · W 1 )| stays above ε ∈ (0,1] in the long run decays to zero exponentially fast ~ e -kI . Furthermore, we explicitly characterize the rate of this convergence I and show that it depends only on the underlying graphs that support the matrices W k ‘s. Our results reveal that the rate I is essentially determined by the most likely way in which the union (over time) of the support graphs fails to form a directed tree.

Products of stochastic matrices: Large deviation rate for Markov chain temporal dependencies

Bajovic D., Xavier J., Sinopoli B.

2012 50th Annual Allerton Conference on Communication, Control, and Computing, Allerton 2012

2012

Abstract:

We find the large deviation rate I for convergence in probability of the product W k —W 1 W 0 of temporally dependent random stochastic matrices. As the model for temporal dependencies, we adopt the Markov chain whose set of states is the set of all possible graphs that support the matrices W k . Such model includes, for example, 1) token-based protocols, where a token is passed among nodes to determine the order of processing; and 2) temporally dependent link failures, where the temporal dependence is modeled by a Markov chain. We characterize the rate I as a function of the Markov chain transition probability matrix P. Examples further reveal how the temporal correlations (dependencies) affect the rate of convergence in probability I.

Project INSIDE: towards autonomous semi-unstructured human–robot social interaction in autism therapy

Melo F.S., Sardinha A., Belo D., Couto M., Faria M., Farias A., Gambôa H., Jesus C., Kinarullathil M., Lima P., Luz L., Mateus A., Melo I., Moreno P., Osório D., Paiva A., Pimentel J., Rodrigues J., Ventura R.

Artificial Intelligence in Medicine

2018

Abstract:

This paper describes the INSIDE system, a networked robot system designed to allow the use of mobile robots as active players in the therapy of children with autism spectrum disorders (ASD). While a significant volume of work has explored the impact of robots in ASD therapy, most such work comprises remotely operated robots and/or well-structured interaction dynamics. In contrast, the INSIDE system allows for complex, semi-unstructured interaction in ASD therapy while featuring a fully autonomous robot. In this paper we describe the hardware and software infrastructure that supports such rich form of interaction, as well as the design methodology that guided the development of the INSIDE system. We also present some results on the use of our system both in pilot and in a long-term study comprising multiple therapy sessions with children at Hospital Garcia de Orta, in Portugal, highlighting the robustness and autonomy of the system as a whole.

Project MAIA: Multilingual AI Agent Assistant

Graça J., Dimas P., Moniz H., Martins A., Neubig G

EAMT 2020

2020

Abstract:

This paper presents the Multilingual Artificial Intelligence Agent Assistant (MAIA), a project led by Unbabel with the collaboration of CMU, INESC-ID and IT Lisbon. MAIA will employ cutting-edge machine learning and natural language processing technologies to build multilingual AI agent assistants, eliminating language barriers. MAIA’s translation layer will empower human agents to provide customer support in real-time, in any language, with human quality

Project MAIA: Multilingual AI Agent Assistant

Martins A., Graça J., Dimas P., Moniz H., Neubig G.

Proceedings of the 22nd Annual Conference of the European Association for Machine Translation

2020

Abstract:

This paper presents the Multilingual Artificial Intelligence Agent Assistant (MAIA), a project led by Unbabel with the collaboration of CMU, INESC-ID and IT Lisbon. MAIA will employ cutting-edge machine learning and natural language processing technologies to build multilingual AI agent assistants, eliminating language barriers. MAIA’s translation layer will empower human agents to provide customer support in real-time, in any language, with human quality.

Promoting Fairness through Hyperparameter Optimization

Cruz A., Saleiro P., Belem C., Soares C., Bizarro P.

ICLR'2021

2021

Abstract:

Considerable research effort has been guided towards algorithmic fairness but real-world adoption of bias reduction techniques is still scarce. Existing methods are either metric- or model-specific, require access to sensitive attributes at inference time, or carry high development and deployment costs. This work explores, in the context of a real-world fraud detection application, the unfairness that emerges from traditional ML model development, and how to mitigate it with a simple and easily deployed intervention: fairness-aware hyperparameter optimization (HO). We propose and evaluate fairness-aware variants of three popular HO algorithms: Fair Random Search, Fair TPE, and Fairband. Our method enables practitioners to adapt pre-existing business operations to accommodate fairness objectives in a frictionless way and with controllable fairness-accuracy trade-offs. Additionally, it can be coupled with existing bias reduction techniques to tune their hyperparameters. We validate our approach on a real-world bank account opening fraud use case, as well as on three datasets from the fairness literature. Results show that, without extra training cost, it is feasible to find models with 111% average fairness increase and just 6% decrease in predictive accuracy, when compared to standard fairness-blind HO.

Promoting Fairness through Hyperparameter Optimization

Cruz A., Saleiro P., Belém C., Soares C., Bizarro P.

arxiv

2021

Abstract:

Considerable research effort has been guided towards algorithmic fairness but real-world adoption of bias reduction techniques is still scarce. Existing methods are either metric- or model-specific, require access to sensitive attributes at inference time, or carry high development and deployment costs. This work explores, in the context of a real-world fraud detection application, the unfairness that emerges from traditional ML model development, and how to mitigate it with a simple and easily deployed intervention: fairness-aware hyperparameter optimization (HO). We propose and evaluate fairness-aware variants of three popular HO algorithms: Fair Random Search, Fair TPE, and Fairband1 . Our method enables practitioners to adapt pre-existing business operations to accommodate fairness objectives in a frictionless way and with controllable fairness-accuracy trade-offs. Additionally, it can be coupled with existing bias reduction techniques to tune their hyperparameters. We validate our approach on a real-world bank account opening fraud use case, as well as on three datasets from the fairness literature. Results show that, without extra training cost, it is feasible to find models with 111% average fairness increase and just 6% decrease in predictive accuracy, when compared to standard fairness-blind HO.

Proof-carrying code in a session-typed process calculus

Pfenning F., Caires L., Toninho B.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

2011

Abstract:

Dependent session types allow us to describe not only properties of the I/O behavior of processes but also of the exchanged data. In this paper we show how to exploit dependent session types to express proof-carrying communication. We further introduce two modal operators into the type theory to provide detailed control about how much information is communicated: one based on traditional proof irrelevance and one integrating digital signatures.