Conference Papers

Rely-guarantee protocols

Militao F., Aldrich J., Caires L.

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

2014

Abstract:

The use of shared mutable state, commonly seen in object-oriented systems, is often problematic due to the potential conflicting interactions between aliases to the same state. We present a substructural type system outfitted with a novel lightweight interference control mechanism, rely-guarantee protocols, that enables controlled aliasing of shared resources. By assigning each alias separate roles, encoded in a novel protocol abstraction in the spirit of rely-guarantee reasoning, our type system ensures that challenging uses of shared state will never interfere in an unsafe fashion. In particular, rely-guarantee protocols ensure that each alias will never observe an unexpected value, or type, when inspecting shared memory regardless of how the changes to that shared state (originating from potentially unknown program contexts) are interleaved at run-time.

Reordering modeling using weighted alignment matrices

Ling W., Luis T., Graca J., Coheur L., Trancoso I.

ACL-HLT 2011 - Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies

2011

Abstract:

In most statistical machine translation systems, the phrase/rule extraction algorithm uses alignments in the 1-best form, which might contain spurious alignment points. The usage of weighted alignment matrices that encode all possible alignments has been shown to generate better phrase tables for phrase-based systems. We propose two algorithms to generate the well known MSD reordering model using weighted alignment matrices. Experiments on the IWSLT 2010 evaluation datasets for two language pairs with different alignment algorithms show that our methods produce more accurate reordering models, as can be shown by an increase over the regular MSD models of 0.4 BLEU points in the BTEC French to English test set, and of 1.5 BLEU points in the DIALOG Chinese to English test set.

Resource allocation in the network operator’s cloud: A virtualization approach

Soares J., Carapinha J., Melo M., Monteiro R., Sargento S.

Proceedings - IEEE Symposium on Computers and Communications

2012

Abstract:

The access infrastructure to the Cloud is usually a major drawback that limits the uptake of Cloud services. Attention has turned to rethinking a new architectural deployment of the overall Cloud service delivery. We argue that it is not sufficient to integrate the cloud domain with the operator’s network domain based on the current models. In this work we envision a full integration of the Cloud and the network, where cloud resources are no longer confined to a data center, but are spread throughout the network and owned by the network operator. In such an environment, challenges arise at different levels, such as at the resource management, where both cloud and network resources need to be managed in an integrated approach. We particularly address the resource allocation problem through joint virtualization of network and cloud resources, by proposing an algorithm to allocate cloud and network resources in an integrated way. This algorithm is evaluated through both simulation and experimental results in a real virtualization platform.

Robotic Versus Human Coaches for Active Aging: An Automated Social Presence Perspective

Caic M., Avelino J., Mahr D., Odekerken G., Odekerken-Schroder G., Bernardino A.

International Journal of Social Robotics

2019

Abstract:

This empirical study compares elderly people’s social perception of human versus robotic coaches in the context of an active and healthy ageing program. In evaluating hedonic and utilitarian value perceptions of exergames (i.e., video games integrating physical activity), we consider elderly people’s judgments of warmth and competence (i.e., social cognition) of their assigned coach (human vs. robot). Our field experiments involve 58 elderly participants in the real-life context. Leveraging a mixed-method approach, combining quantitative and qualitative data, we show that i) socially assistive robots activate feelings of (automated) social presence in the elderly; ii) human coaches score higher on perceived warmth and competence relative to robotic coaches; iii) social cognition impacts elderly people’s

Robust and speculative byzantine randomized consensus with constant time complexity in normal conditions

Vavala B., Neves N.

Proceedings of the IEEE Symposium on Reliable Distributed Systems

2012

Abstract:

Randomized Byzantine Consensus can be an interesting building block in the implementation of asynchronous distributed systems. Despite its exponential worst-case complexity, which would make it less appealing in practice, a few experimental works have argued quite the opposite. To bridge the gap between theory and practice, we analyze a well-known state-of-the-art algorithm in normal system conditions, in which crash failures may occur but no malicious attacks, proving that it is fast on average. We then leverage our analysis to improve its best-case complexity from three to two phases, by reducing the communication operations through speculative executions. Our findings are confirmed through an experimental validation.

Robust linear dimensionality reduction for hypothesis testing with application to sensor selection

Bajovic D., Sinopoli B., Xavier J.

2009 47th Annual Allerton Conference on Communication, Control, and Computing, Allerton 2009

2009

Abstract:

This paper addresses robust linear dimensionality reduction (RLDR) for binary Gaussian hypothesis testing. The goal is to find a linear map from the high dimensional space where the data vector lives to a low dimensional space where the hypothesis test is carried out. The linear map is designed to maximize the detector performance. This translates into maximizing the Kullback-Leibler (KL) distance between the two projected distributions. In practice, the distribution parameters are estimated from training data, thus subject to uncertainty. This is modeled by allowing the distribution parameters to drift within some confidence regions. We address the case where only the mean values of the Gaussian distributions, m0 and m1, are uncertain with confidence ellipsoids defined by the corresponding covariance matrices, S0 and S1. Under this setup, we find the linear map that maximizes the KL distance for the worst case drift of the mean values. We solve the problem globally for the case of linear mapping to one dimension, reducing it to a grid search over a finite interval. Our solution shows superior performance compared to robust linear discriminant analysis techniques recently proposed in the literature. In addition, we use our RLDR solution as a building block to derive a sensor selection algorithm for robust event detection, in the context of sensor networks. Our sensor selection algorithm shows quasi-optimal performance: worst-case KL distance for suboptimal sensor selection is at most 15% smaller than worst-case KL distance for the optimal sensor selection obtained by exhaustive search.

Robust Object Recognition Through Symbiotic Deep Learning In Mobile Robots*

Cartucho J., Ventura R., Veloso M.

Proc. of IROS 2018 - IEEE/RSJ International Conference on Intelligent Robots and Systems, Madrid, Spain

2018

Abstract:

Despite the recent success of state-of-the-art deep learning algorithms in object recognition, when these are deployed as-is on a mobile service robot, we observed that they failed to recognize many objects in real human environments. In this paper, we introduce a learning algorithm in which robots address this flaw by asking humans for help, also known as symbiotic autonomy approach. In particular, we bootstrap YOLOv2, a state-of-the-art deep neural network and create a HUMAN neural net using only the collected data. Using an RGB camera and an on-board tablet, the robot proactively seeks for human input to assist in labeling surrounding objects. Pepper, based in CMU, and Monarch Mbot, based in ISR-Lisbon, are the social robots that we used to validate the proposed approach. We conducted a study in a realistic domestic environment over the course of 20 days with 6 research participants. To improve object recognition, we used the two neural nets, YOLOv2 + HUMAN, in parallel. The robot collects data about where an object is and to whom it belongs by asking. This enabled us to introduce an approach where the robot can search for a specific person’s object. We view the contribution of this paper to be relevant for service robots in general, in addition to Pepper and Mbot. Following this methodology, the robot was able to detect twice the number of objects compared to the initial YOLOv2, with an improved average precision.

Robust Regression

Huang D., Cabral R., Torre F.D.

IEEE Transactions on Pattern Analysis and Machine Intelligence

2016

Abstract:

Discriminative methods (e.g., kernel regression, SVM) have been extensively used to solve problems such as object recognition, image alignment and pose estimation from images. These methods typically map image features (X) to continuous (e.g., pose) or discrete (e.g., object category) values. A major drawback of existing discriminative methods is that samples are directly projected onto a subspace and hence fail to account for outliers common in realistic training sets due to occlusion, specular reflections or noise. It is important to notice that existing discriminative approaches assume the input variables X to be noise free. Thus, discriminative methods experience significant performance degradation when gross outliers are present. Despite its obvious importance, the problem of robust discriminative learning has been relatively unexplored in computer vision. This paper develops the theory of robust regression (RR) and presents an effective convex approach that uses recent advances on rank minimization. The framework applies to a variety of problems in computer vision including robust linear discriminant analysis, regression with missing data, and multi-label classification. Several synthetic and real examples with applications to head pose estimation from images, image and video classification and facial attribute classification with missing data are used to illustrate the benefits of RR.

Say yes to the fetch: Product retrieval on a structured multimodal catalog

Morais J., Santiago C., Costeira J.P.

1st Annual AAAI Workshop on AI to Accelerate Science and Engineering

2020

Abstract:

Multimodal conversational agents allow the user to communicate through natural language and visual information. In ecommerce, this type of agents have the potential to lead to realistic and dynamic shopping experiences, where the costumer finds the desired products more efficiently with the help of an agent. A common approach in this scenario is to build a representation space where both the textual and visual information of a product are close. Then, it is possible to search and retrieve products with queries from any of the modalities. This work proposes to generate this joint representation space by also taking into account prior knowledge about the fashion domain, to ensure that the retrieved products comply with the target type of products. Combining label relaxation with a taxonomy-based regularization, the proposed approach diminishes the penalization of the contrastive loss by assigning a smaller loss to other acceptable matches. Our results show that the proposed approach significantly reduces gross errors, like retrieving pants when the costumer is looking for t-shirts, while simultaneously achieving good retrieval performances. Additionally, this approach allows multimodal queries, where specific attributes can be modified by manipulating a visual query with text.