Le traitement du langage naturel à l'ère de l'IA générative
Rejoignez-nous à Mila en octobre pour un atelier de trois jour visant à explorer le potentiel de transformation des technologies langagières et leurs implications pour la société.
Ce programme est conçu pour fournir aux professionnel·le·s travaillant dans le domaine de la politique une compréhension fondamentale de la technologie de l'IA.
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.) ?
. Column generation is a popular method to solve large-scale linear programs with an exponential number of variables. Several important appl… (voir plus)ications, such as the vehicle routing problem, rely on this technique in order to be solved. However, in practice, column generation methods suffer from slow convergence (i.e. they require too many iterations). Stabilization techniques, which carefully select the column to add at each iteration, are commonly used to improve convergence. In this work, we frame the problem of selecting which columns to add as one of sequential decision-making. We propose a neural column generation architecture that iteratively selects columns to be added to the problem. Our architecture is inspired by stabilization techniques and predicts the optimal duals, which are then used to select the columns to add. We proposed architecture, trained using imitation learning. Exemplified on the Vehicle Routing Problem, we show that several machine learning models yield good performance in predicting the optimal duals and that our architecture outperforms them as well as a popular state-of-the-art stabilization technique. Further, the architecture approach can generalize to instances larger than those observed during training.
Multi-head, key-value attention is the backbone of transformer-like model architectures which have proven to be widely successful in recent … (voir plus)years. This attention mechanism uses multiple parallel key-value attention blocks (called heads), each performing two fundamental computations: (1) search - selection of a relevant entity from a set via query-key interaction, and (2) retrieval - extraction of relevant features from the selected entity via a value matrix. Standard attention heads learn a rigid mapping between search and retrieval. In this work, we first highlight how this static nature of the pairing can potentially: (a) lead to learning of redundant parameters in certain tasks, and (b) hinder generalization. To alleviate this problem, we propose a novel attention mechanism, called Compositional Attention, that replaces the standard head structure. The proposed mechanism disentangles search and retrieval and composes them in a dynamic, flexible and context-dependent manner. Through a series of numerical experiments, we show that it outperforms standard multi-head attention on a variety of tasks, including some out-of-distribution settings. Through our qualitative analysis, we demonstrate that Compositional Attention leads to dynamic specialization based on the type of retrieval needed. Our proposed mechanism generalizes multi-head attention, allows independent scaling of search and retrieval and is easy to implement in a variety of established network architectures.
2022-01-01
International Conference on Learning Representations (publié)
The mixing time of the Markov chain induced by a policy limits performance in real-world continual learning scenarios. Yet, the effect of mi… (voir plus)xing times on learning in continual reinforcement learning (RL) remains underexplored. In this paper, we characterize problems that are of long-term interest to the development of continual RL, which we call scalable MDPs, through the lens of mixing times. In particular, we theoretically establish that scalable MDPs have mixing times that scale polynomially with the size of the problem. We go on to demonstrate that polynomial mixing times present significant difficulties for existing approaches that suffer from myopic bias and stale bootstrapped estimates. To validate the proposed theory, we study the empirical scaling behavior of mixing times with respect to the number of tasks and task switching frequency for pretrained high performing policies on seven Atari games. Our analysis demonstrates both that polynomial mixing times do emerge in practice and how their existence may lead to unstable learning behavior like catastrophic forgetting in continual learning settings.
This work introduces a novel principle we call disentanglement via mechanism sparsity regularization, which can be applied when the latent f… (voir plus)actors of interest depend sparsely on past latent factors and/or observed auxiliary variables. We propose a representation learning method that induces disentanglement by simultaneously learning the latent factors and the sparse causal graphical model that relates them. We develop a rigorous identifiability theory, building on recent nonlinear independent component analysis (ICA) results, that formalizes this principle and shows how the latent variables can be recovered up to permutation if one regularizes the latent mechanisms to be sparse and if some graph connectivity criterion is satisfied by the data generating process. As a special case of our framework, we show how one can leverage unknown-target interventions on the latent factors to disentangle them, thereby drawing further connections between ICA and causality. We propose a VAE-based method in which the latent mechanisms are learned and regularized via binary masks, and validate our theory by showing it learns disentangled representations in simulations.
As the number of antennas increases in multi-input and multi-output (MIMO) systems, even linear detection methods suffer from sharply increa… (voir plus)sing complexity. This paper proposes a learning-based multi-layer perception (MLP), named dynamic stochastic multi-layer perception (DsMLP), which is implemented by dynamic stochastic computing (DSC). We first establish a similar form between the MLP structure and minimum mean square error (MMSE) matrix operations. Consequently, DsMLP transforms the complex computation problem into an optimization problem of MLP training. Due to the specific design of MLP structure, e.g., same input/output dimension and single layer without activation function, the mathematical representation of DsMLP is identical to the MMSE matrix operations. Therefore, DsMLP guarantees sound model explainability in mathematics, fast convergence in training, and low complexity in computation. Furthermore, we transform the MLP training process to the DSC domain and propose a hardware-efficient scheme for DsMLP. Compared with other state-of-the-art MIMO detectors, DsMLP achieves 1.2× energy efficiency and 1.74× area efficiency.
As the number of antennas increases in multi-input and multi-output (MIMO) systems, even linear detection methods suffer from sharply increa… (voir plus)sing complexity. This paper proposes a learning-based multi-layer perception (MLP), named dynamic stochastic multi-layer perception (DsMLP), which is implemented by dynamic stochastic computing (DSC). We first establish a similar form between the MLP structure and minimum mean square error (MMSE) matrix operations. Consequently, DsMLP transforms the complex computation problem into an optimization problem of MLP training. Due to the specific design of MLP structure, e.g., same input/output dimension and single layer without activation function, the mathematical representation of DsMLP is identical to the MMSE matrix operations. Therefore, DsMLP guarantees sound model explainability in mathematics, fast convergence in training, and low complexity in computation. Furthermore, we transform the MLP training process to the DSC domain and propose a hardware-efficient scheme for DsMLP. Compared with other state-of-the-art MIMO detectors, DsMLP achieves 1.2× energy efficiency and 1.74× area efficiency.
2022-01-01
IEEE Transactions on Signal Processing (published)
Online escort advertisement websites are widely used for advertising victims of human trafficking. Domain experts agree that advertising mul… (voir plus)tiple people in the same ad is a strong indicator of trafficking. Thus, extracting person names from the text of these ads can provide valuable clues for further analysis. However, Named-Entity Recognition (NER) on escort ads is challenging because the text can be noisy, colloquial and often lacking proper grammar and punctuation. Most existing state-of-the-art NER models fail to demonstrate satisfactory performance in this task. In this paper, we propose NEAT (Name Extraction Against Trafficking) for extracting person names. It effectively combines classic rule-based and dictionary extractors with a contextualized language model to capture ambiguous names (e.g penny, hazel) and adapts to adversarial changes in the text by expanding its dictionary. NEAT shows 19% improvement on average in the F1 classification score for name extraction compared to previous state-of-the-art in two domain-specific datasets.
The surging demand for multilingual dialogue systems often requires a costly labeling process for each language addition. For low resource l… (voir plus)anguages, human annotators are continuously tasked with the adaptation of resource-rich language utterances for each new domain. However, this prohibitive and impractical process can often be a bottleneck for low resource languages that are still without proper translation systems nor parallel corpus. In particular, it is difficult to obtain task-specific low resource language annotations for the English-derived creoles (e.g. Nigerian and Cameroonian Pidgin). To address this issue, we utilize the pretrained language models i.e. BART which has shown great potential in language generation/understanding – we propose to finetune the BART model to generate utterances in Pidgin by leveraging the proximity of the source and target languages, and utilizing positive and negative examples in constrastive training objectives. We collected and released the first parallel Pidgin-English conversation corpus in two dialogue domains and showed that this simple and effective technique is suffice to yield impressive results for English-to-Pidgin generation, which are two closely-related languages.