Conference Papers

Nonlinear estimation using Mean Field Games

Pequito S., Aguiar A.P., Sinopoli B., Gomes D.A.

International Conference on NETwork Games, Control and Optimization, NetGCooP 2011

2011

Abstract:

This paper introduces Mean Field Games (MFG) as a framework to develop optimal estimators in some sense for a general class of nonlinear systems. We show that under suitable conditions the estimation error converges exponentially fast to zero. Computer simulations are performed to illustrate the method. In particular we provide an example where the proposed estimator converges whereas both extended Kalman filter and particle filter diverge.

Not all contexts are created equal: Better word representations with variable attention

Ling W., Chu-Cheng L., Tsvetkov Y., Amir S., Astudillo R.F., Dyer C., Black A.W., Trancoso I.

Conference Proceedings - EMNLP 2015: Conference on Empirical Methods in Natural Language Processing

2015

Abstract:

We introduce an extension to the bag-ofwords model for learning words representations that take into account both syntactic and semantic properties within language. This is done by employing an attention model that finds within the contextual words, the words that are relevant for each prediction. The general intuition of our model is that some words are only relevant for predicting local context (e.g. function words), while other words are more suited for determining global context, such as the topic of the document. Experiments performed on both semantically and syntactically oriented tasks show gains using our model over the existing bag of words model. Furthermore, compared to other more sophisticated models, our model scales better as we increase the size of the context of the model.

Numerical Methods for Finite-State Mean-Field Games Satisfying a Monotonicity Condition

Gomes A.D.; Saúde J.

Applied Mathematics & Optimization

2018

Abstract:

Here, we develop numerical methods for finite-state mean-field games (MFGs) that satisfy a monotonicity condition. MFGs are determined by a system of differential equations with initial and terminal boundary conditions. These non-standard conditions make the numerical approximation of MFGs difficult. Using the monotonicity condition, we build a flow that is a contraction and whose fixed points solve both for stationary and time-dependent MFGs. We illustrate our methods with a MFG that models the paradigm-shift problem.

O‐MedAL: Online active deep learning for medical image analysis

Smailagic, A. ; Costa, P. ;Gaudio, A. ; Khandelwal, K. ; Mirshekari, M. ; Fagert, J. ; Walawalkar, D. ; Xu, S. ; Galdran, A. ; Zhang, P. ; Campilho, A. ; Noh, H.

WIREs Data Mining and Knowledge Discovery

2020

Abstract:

Active learning (AL) methods create an optimized labeled training set from unlabeled data. We introduce a novel online active deep learning method for medical image analysis. We extend our MedAL AL framework to present new results in this paper. A novel sampling method queries the unlabeled examples that maximize the average distance to all training set examples. Our online method enhances performance of its underlying baseline deep network. These novelties contribute to significant performance improvements, including improving the model’s underlying deep network accuracy by 6.30%, using only 25% of the labeled dataset to achieve baseline accuracy, reducing backpropagated images during training by as much as 67%, and demonstrating robustness to class imbalance in binary and multiclass tasks. This article is categorized under: Technologies > Machine Learning Technologies > Classification Application Areas > Health Care

On asynchronous implementations of fictitious play for distributed learning

Swenson B., Kar S., Xavier J.

Conference Record - Asilomar Conference on Signals, Systems and Computers

2016

Abstract:

The classical Fictitious Play (FP) algorithm is defined within a framework of synchronous repeated play. In practice, the global synchronization assumed in classical FP can be difficult to achieve in large-scale multi-agent settings. The paper considers FP with asynchronous updates-a variant of FP in which players are permitted to be either “active” or “idle” in each stage of the repeated play process. The FP process with asynchronous updates is shown to be a generalization of classical FP. Analytical convergence results are given for the asynchronous variant of FP. Furthermore, the paper studies an asynchronous continuous-time embedding of FP. The continuous- time embedded FP process may be implemented in a real-world setting where no global clock is available. Sufficient conditions for convergence of the continuous-time embedded process are provided as a consequence of the convergence analysis for FP with asynchronous updates. Example implementations that attain the sufficient condition are presented.

On compiling linear logic programs with comprehensions, aggregates and rule priorities

Cruz F., Rocha R.

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

2015

Abstract:

Linear logic programs are challenging to implement efficiently because facts are asserted and retracted frequently. Implementation is made more difficult with the introduction of useful features such as rule priorities, which are used to specify the order of rule inference, and comprehensions or aggregates, which are mechanisms that make data iteration and gathering more intuitive. In this paper, we describe a compilation scheme for transforming linear logic programs enhanced with those features into efficient C++ code. Our experimental results show that compiled logic programs are less than one order of magnitude slower than hand-written C programs and much faster than interpreted languages such as Python.

On the complexity of the constrained input selection problem for structural linear systems

Pequito S., Kar S., Aguiar A.P.

Automatica

2015

Abstract:

This paper studies the problem of, given the structure of a linear-time invariant system and a set of possible inputs, finding the smallest subset of input vectors that ensures system’s structural controllability. We refer to this problem as the minimum constrained input selection (minCIS) problem, since the selection has to be performed on an initial given set of possible inputs. We prove that the minCIS problem is NP-hard, which addresses a recent open question of whether there exist polynomial algorithms (in the size of the system plant matrices) that solve the minCIS problem. To this end, we show that the associated decision problem, to be referred to as the CIS, of determining whether a subset (of a given collection of inputs) with a prescribed cardinality exists that ensures structural controllability, is NP-complete. Further, we explore in detail practically important subclasses of the minCIS obtained by introducing more specific assumptions either on the system dynamics or the input set instances for which systematic solution methods are provided by constructing explicit reductions to well known computational problems. The analytical findings are illustrated through examples in multi-agent leader–follower type control problems.

On the connection availability between relay nodes in a VANET

Cardote A., Sargento S., Steenkiste P.

2010 IEEE Globecom Workshops, GC'10

2010

Abstract:

Vehicular Ad-hoc NETworks (VANET) are emerging as a promising way to disseminate information among vehicles. As this information can range from safety to infotainment application content, and the vehicular environment has very particular characteristics, the behavior of the network must be effectively studied in order to adapt the transmission mechanisms. This work presents a model for the connectivity patterns of chains of vehicles traveling in a highway. This information will be crucial to provide insight in the design of VANET protocols and applications, which will be dependent on the connectivity characteristics. The accuracy of the model is shown through its application to specific case studies. The obtained results show that, in highway scenarios, the connectivity availability between relay nodes can last for a significant amount of time (in the order of tens of seconds).

On the convergence of finite state mean-field games through Γ-convergence

Ferreira R., Gomes D.A.

Journal of Mathematical Analysis and Applications

2014

Abstract:

In this study, we consider the long-term convergence (trend toward an equilibrium) of finite state mean-field games using Γ-convergence. Our techniques are based on the observation that an important class of mean-field games can be viewed as the Euler–Lagrange equation of a suitable functional. Therefore, using a scaling argument, one can convert a long-term convergence problem into a Γ-convergence problem. Our results generalize previous results related to long-term convergence for finite state problems.