Author name cluster

André Martins

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

4 papers

2 author rows

NeurIPS Conference 2022 Conference Paper

Learning to Scaffold: Optimizing Model Explanations for Teaching

Patrick Fernandes
Marcos Treviso
Danish Pruthi
André Martins
Graham Neubig

Modern machine learning models are opaque, and as a result there is a burgeoning academic subfield on methods that explain these models' behavior. However, what is the precise goal of providing such explanations, and how can we demonstrate that explanations achieve this goal? Some research argues that explanations should help teach a student (either human or machine) to simulate the model being explained, and that the quality of explanations can be measured by the simulation accuracy of students on unexplained examples. In this work, leveraging meta-learning techniques, we extend this idea to improve the quality of the explanations themselves, specifically by optimizing explanations such that student models more effectively learn to simulate the original model. We train models on three natural language processing and computer vision tasks, and find that students trained with explanations extracted with our framework are able to simulate the teacher significantly more effectively than ones produced with previous methods. Through human annotations and a user study, we further find that these learned explanations more closely align with how humans would explain the required decisions in these tasks. Our code is available at https: //github. com/coderpat/learning-scaffold.

PDF Details

NeurIPS Conference 2020 Conference Paper

Efficient Marginalization of Discrete and Structured Latent Variables via Sparsity

Gonçalo Correia
Vlad Niculae
Wilker Aziz
André Martins

Training neural network models with discrete (categorical or structured) latent variables can be computationally challenging, due to the need for marginalization over large or combinatorial sets. To circumvent this issue, one typically resorts to sampling-based approximations of the true marginal, requiring noisy gradient estimators (e. g. , score function estimator) or continuous relaxations with lower-variance reparameterized gradients (e. g. , Gumbel-Softmax). In this paper, we propose a new training strategy which replaces these estimators by an exact yet efficient marginalization. To achieve this, we parameterize discrete distributions over latent assignments using differentiable sparse mappings: sparsemax and its structured counterparts. In effect, the support of these distributions is greatly reduced, which enables efficient marginalization. We report successful results in three tasks covering a range of latent variable modeling applications: a semisupervised deep generative model, a latent communication game, and a generative model with a bit-vector latent representation. In all cases, we obtain good performance while still achieving the practicality of sampling-based approximations.

PDF Details

NeurIPS Conference 2020 Conference Paper

Sparse and Continuous Attention Mechanisms

André Martins
António Farinhas
Marcos Treviso
Vlad Niculae
Pedro Aguiar
Mario Figueiredo

Exponential families are widely used in machine learning; they include many distributions in continuous and discrete domains (e. g. , Gaussian, Dirichlet, Poisson, and categorical distributions via the softmax transformation). Distributions in each of these families have fixed support. In contrast, for finite domains, there has been recent work on sparse alternatives to softmax (e. g. , sparsemax and alpha-entmax), which have varying support, being able to assign zero probability to irrelevant categories. These discrete sparse mappings have been used for improving interpretability of neural attention mechanisms. This paper expands that work in two directions: first, we extend alpha-entmax to continuous domains, revealing a link with Tsallis statistics and deformed exponential families. Second, we introduce continuous-domain attention mechanisms, deriving efficient gradient backpropagation algorithms for alpha in {1, 2}. Experiments on attention-based text classification, machine translation, and visual question answering illustrate the use of continuous attention in 1D and 2D, showing that it allows attending to time intervals and compact regions.

PDF Details

IROS Conference 2005 Conference Paper

Environmental monitoring with mobile robots

Lino Marques
André Martins
Aníbal T. de Almeida

This paper describes an architecture for estimating environmental odor maps and presents experimental results obtained using that architecture with five mobile robots inside a large laboratory with two odor sources and forced ventilation. The mobile sensing agents employed in the experiments have self localization capabilities and carry an electronic nose and a thermal anemometer. The proposed architecture allows integrating sparse olfaction data gathered by the mobile agents along their trajectories and dynamically estimate the spatial concentration of different odor fields. The data assimilation process is made centrally by a PC that polls periodically each robot through a RF network in order to get the data gathered during the previous acquisition period. The estimation of odor fields is made in two steps: first each agent estimates the odor mixture and concentration by means of a neural network-based regression algorithm that converts values from gas sensor space to the corresponding odor space. Then the sensed data is assimilated into an advection-diffusion model by means of a reduced order Kalman filter. In the current implementation the central controller, responsible for the maps estimation, specifies to each agent their target area to explore. The proposed architecture was validated with a set of experiments that demonstrated its ability to estimate and capture the dynamics of overlapping odor fields. This work can easily be adapted to city buses or other local transportation systems in order to monitor the pollution or quickly detect hazardous chemicals inside cities.

Details