Nous utilisons des témoins pour analyser le trafic et l’utilisation de notre site web, afin de personnaliser votre expérience. Vous pouvez désactiver ces technologies à tout moment, mais cela peut restreindre certaines fonctionnalités du site. Consultez notre Politique de protection de la vie privée pour en savoir plus.
Paramètre des cookies
Vous pouvez activer et désactiver les types de cookies que vous souhaitez accepter. Cependant certains choix que vous ferez pourraient affecter les services proposés sur nos sites (ex : suggestions, annonces personnalisées, etc.).
Cookies essentiels
Ces cookies sont nécessaires au fonctionnement du site et ne peuvent être désactivés. (Toujours actif)
Cookies analyse
Acceptez-vous l'utilisation de cookies pour mesurer l'audience de nos sites ?
Multimedia Player
Acceptez-vous l'utilisation de cookies pour afficher et vous permettre de regarder les contenus vidéo hébergés par nos partenaires (YouTube, etc.) ?
Publications
Meta-learning framework with applications to zero-shot time-series forecasting
Can meta-learning discover generic ways of processing time series (TS) from a diverse dataset so as to greatly improve generalization on new… (voir plus) TS coming from different datasets? This work provides positive evidence to this using a broad meta-learning framework which we show subsumes many existing meta-learning algorithms. Our theoretical analysis suggests that residual connections act as a meta-learning adaptation mechanism, generating a subset of task-specific parameters based on a given TS input, thus gradually expanding the expressive power of the architecture on-the-fly. The same mechanism is shown via linearization analysis to have the interpretation of a sequential update of the final linear layer. Our empirical results on a wide range of data emphasize the importance of the identified meta-learning mechanisms for successful zero-shot univariate forecasting, suggesting that it is viable to train a neural network on a source TS dataset and deploy it on a different target TS dataset without retraining, resulting in performance that is at least as good as that of state-of-practice univariate forecasting models.
2021-05-18
Proceedings of the AAAI Conference on Artificial Intelligence (publié)
Branch and Bound (B&B) is the exact tree search method typically used to solve Mixed-Integer Linear Programming problems (MILPs). Learning b… (voir plus)ranching policies for MILP has become an active research area, with most works proposing to imitate the strong branching rule and specialize it to distinct classes of problems. We aim instead at learning a policy that generalizes across heterogeneous MILPs: our main hypothesis is that parameterizing the state of the B&B search tree can aid this type of generalization. We propose a novel imitation learning framework, and introduce new input features and architectures to represent branching. Experiments on MILP benchmark instances clearly show the advantages of incorporating an explicit parameterization of the state of the search tree to modulate the branching decisions, in terms of both higher accuracy and smaller B&B trees. The resulting policies significantly outperform the current state-of-the-art method for "learning to branch" by effectively allowing generalization to generic unseen instances.
2021-05-18
Proceedings of the AAAI Conference on Artificial Intelligence (publié)
Visual saliency has emerged as a major visualization tool for interpreting deep reinforcement learning (RL) agents. However, much of the exi… (voir plus)sting research uses it as an analyzing tool rather than an inductive bias for policy learning. In this work, we use visual attention as an inductive bias for RL agents. We propose a novel self-supervised attention learning approach which can 1. learn to select regions of interest without explicit annotations, and 2. act as a plug for existing deep RL methods to improve the learning performance. We empirically show that the self-supervised attention-aware deep RL methods outperform the baselines in the context of both the rate of convergence and performance. Furthermore, the proposed self-supervised attention is not tied with specific policies, nor restricted to a specific scene. We posit that the proposed approach is a general self-supervised attention module for multi-task learning and transfer learning, and empirically validate the generalization ability of the proposed method. Finally, we show that our method learns meaningful object keypoints highlighting improvements both qualitatively and quantitatively.
2021-05-18
AAAI Conference on Artificial Intelligence (publié)
Jumana Sara Tobias Carsten Michael Daniel Claudia Yvette Bhismadev Chris Ineke Daisy Flavio Jessica Vincent Pilar David Lindsay Hannah Joerg Rosemary J. Xavier Liogier David J. René Andre Maarten Nico Bethany Laurence Bob Gahan Antonio M. Barbara Amber N. V. Jessica Roberto Antonia San José Emily Roberto Heike Jack Steve C. R. Caroline Marcel P. Ahmad
Background: The initial injury burden from incident TBI is significantly amplified by recurrent TBI (rTBI). Unfortunately, research assessin… (voir plus)g the accuracy to conduct rTBI surveillance is not available. Accurate surveillance information on recurrent injuries is needed to justify the allocation of resources to rTBI prevention and to conduct high quality epidemiological research on interventions that mitigate this injury burden. This study evaluates the accuracy of administrative health data (AHD) surveillance case definitions for rTBI and estimates the 1-year rTBI incidence adjusted for measurement error. Methods: A 25% random sample of AHD for Montreal residents from 2000 to 2014 was used in this study. Four widely used TBI surveillance case definitions, based on the International Classification of Disease and on radiological exams of the head, were applied to ascertain suspected rTBI cases. Bayesian latent class models were used to estimate the accuracy of each case definition and the 1-year rTBI measurement-error-adjusted incidence without relying on a gold standard rTBI definition that does not exist, across children (18 years), adults (18-64 years), and elderly (> =65 years). Results: The adjusted 1-year rTBI incidence was 4.48 (95% CrI 3.42, 6.20) per 100 person-years across all age groups, as opposed to a crude estimate of 8.03 (95% CrI 7.86, 8.21) per 100 person-years. Patients with higher severity index TBI had a significantly higher incidence of rTBI compared to patients with lower severity index TBI. The case definition that identified patients undergoing a radiological examination of the head in the context of any traumatic injury was the most sensitive across children [0.46 (95% CrI 0.33, 0.61)], adults [0.79 (95% CrI 0.64, 0.94)], and elderly [0.87 (95% CrI 0.78, 0.95)]. The most specific case definition was the discharge abstract database in children [0.99 (95% CrI 0.99, 1.00)], and emergency room visits claims in adults/elderly [0.99 (95% CrI 0.99, 0.99)]. Median time to rTBI was the shortest in adults (75 days) and the longest in children (120 days). Conclusion: Conducting accurate surveillance and valid epidemiological research for rTBI using AHD is feasible when measurement error is accounted for.