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.) ?
Learning-based approaches for semantic segmentation have two inherent challenges. First, acquiring pixel-wise labels is expensive and time-c… (voir plus)onsuming. Second, realistic segmentation datasets are highly unbalanced: some categories are much more abundant than others, biasing the performance to the most represented ones. In this paper, we are interested in focusing human labelling effort on a small subset of a larger pool of data, minimizing this effort while maximizing performance of a segmentation model on a hold-out set. We present a new active learning strategy for semantic segmentation based on deep reinforcement learning (RL). An agent learns a policy to select a subset of small informative image regions -- opposed to entire images -- to be labeled, from a pool of unlabeled data. The region selection decision is made based on predictions and uncertainties of the segmentation model being trained. Our method proposes a new modification of the deep Q-network (DQN) formulation for active learning, adapting it to the large-scale nature of semantic segmentation problems. We test the proof of concept in CamVid and provide results in the large-scale dataset Cityscapes. On Cityscapes, our deep RL region-based DQN approach requires roughly 30% less additional labeled data than our most competitive baseline to reach the same performance. Moreover, we find that our method asks for more labels of under-represented categories compared to the baselines, improving their performance and helping to mitigate class imbalance.
Reinforcement learning agents that operate in diverse and complex environments can benefit from the structured decomposition of their behavi… (voir plus)or. Often, this is addressed in the context of hierarchical reinforcement learning, where the aim is to decompose a policy into lower-level primitives or options, and a higher-level meta-policy that triggers the appropriate behaviors for a given situation. However, the meta-policy must still produce appropriate decisions in all states. In this work, we propose a policy design that decomposes into primitives, similarly to hierarchical reinforcement learning, but without a high-level meta-policy. Instead, each primitive can decide for themselves whether they wish to act in the current state. We use an information-theoretic mechanism for enabling this decentralized decision: each primitive chooses how much information it needs about the current state to make a decision and the primitive that requests the most information about the current state acts in the world. The primitives are regularized to use as little information as possible, which leads to natural competition and specialization. We experimentally demonstrate that this policy architecture improves over both flat and hierarchical policies in terms of generalization.
In typical Multi-Agent Reinforcement Learning (MARL) settings, each agent acts to maximize its individual reward objective. However, for col… (voir plus)lective social welfare maximization, some agents may need to act non-selfishly. We propose a reward shaping mechanism using extrinsic motivation for achieving modularity and increased cooperation among agents in Sequential Social Dilemma (SSD) problems. Our mechanism, inspired by capitalism, provides extrinsic motivation to agents by redistributing a portion of collected re-wards based on each agent’s individual contribution towards team rewards. We demonstrate empirically that this mechanism leads to higher collective welfare relative to existing baselines. Furthermore, this reduces free rider issues and leads to more diverse policies. We evaluate our proposed mechanism for already specialised agents that are pre-trained for specific roles. We show that our mechanism, in the most challenging CleanUp environment, significantly out-performs two baselines (based roughly on socialism and anarchy) and accumulates 2-3 times higher rewards in an easier setting of the environment.
One of the principal scientific challenges in deep learning is explaining generalization, i.e., why the particular way the community now tra… (voir plus)ins networks to achieve small training error also leads to small error on held-out data from the same population. It is widely appreciated that some worst-case theories -- such as those based on the VC dimension of the class of predictors induced by modern neural network architectures -- are unable to explain empirical performance. A large volume of work aims to close this gap, primarily by developing bounds on generalization error, optimization error, and excess risk. When evaluated empirically, however, most of these bounds are numerically vacuous. Focusing on generalization bounds, this work addresses the question of how to evaluate such bounds empirically. Jiang et al. (2020) recently described a large-scale empirical study aimed at uncovering potential causal relationships between bounds/measures and generalization. Building on their study, we highlight where their proposed methods can obscure failures and successes of generalization measures in explaining generalization. We argue that generalization measures should instead be evaluated within the framework of distributional robustness.
The use of Natural Language (NL) interfaces to allow devices and applications to respond to verbal commands or free-form textual queries is … (voir plus)becoming increasingly prevalent in our society. To a large extent, their success in interpreting and responding to a request is dependent upon rich underlying ontologies and conceptual models that understand the technical or domain specific vocabulary of diverse users. The effective use of NL interfaces in the Software Engineering (SE) domains requires its own ontology models focusing upon software related terms and concepts. While many SE glossaries exist, they are often incomplete and tend to define the vocabulary for specific sub-fields without capturing associations between terms and phrases. This limits their usefulness for supporting NL-related tasks. In this paper we propose an approach for constructing and evolving a semantic network of software engineering concepts and phrases. Our approach starts with a set of existing SE glossaries, uses the existing glossary terms and explicitly defined associations as a starting point, uses machine learning-based techniques to dynamically identify and document additional associations between terms, leverages the network to interpret NL queries in the SE domain, and finally augments the resulting semantic network with feedback provided by users. We evaluate the viability of our approach within the sub-domain of Agile Software Development, focusing on requirements related queries, and show that the semantic network enhances the ability of an NL interface to correctly interpret and execute user queries.
2020-01-01
2020 IEEE Seventh International Workshop on Artificial Intelligence for Requirements Engineering (AIRE) (publié)
Recent work by Brock et al. (2018) suggests that Generative Adversarial Networks (GANs) benefit disproportionately from large mini-batch siz… (voir plus)es. Unfortunately, using large batches is slow and expensive on conventional hardware. Thus, it would be nice if we could generate batches that were effectively large though actually small. In this work, we propose a method to do this, inspired by the use of Coreset-selection in active learning. When training a GAN, we draw a large batch of samples from the prior and then compress that batch using Coreset-selection. To create effectively large batches of 'real' images, we create a cached dataset of Inception activations of each training image, randomly project them down to a smaller dimension, and then use Coreset-selection on those projected activations at training time. We conduct experiments showing that this technique substantially reduces training time and memory usage for modern GAN variants, that it reduces the fraction of dropped modes in a synthetic dataset, and that it allows GANs to reach a new state of the art in anomaly detection.
In artificial neural networks trained with gradient descent, the weights used for processing stimuli are also used during backward passes to… (voir plus) calculate gradients. For the real brain to approximate gradients, gradient information would have to be propagated separately, such that one set of synaptic weights is used for processing and another set is used for backward passes. This produces the so-called "weight transport problem" for biological models of learning, where the backward weights used to calculate gradients need to mirror the forward weights used to process stimuli. This weight transport problem has been considered so hard that popular proposals for biological learning assume that the backward weights are simply random, as in the feedback alignment algorithm. However, such random weights do not appear to work well for large networks. Here we show how the discontinuity introduced in a spiking system can lead to a solution to this problem. The resulting algorithm is a special case of an estimator used for causal inference in econometrics, regression discontinuity design. We show empirically that this algorithm rapidly makes the backward weights approximate the forward weights. As the backward weights become correct, this improves learning performance over feedback alignment on tasks such as Fashion-MNIST, SVHN, CIFAR-10 and VOC. Our results demonstrate that a simple learning rule in a spiking network can allow neurons to produce the right backward connections and thus solve the weight transport problem.
The success of adversarial formulations in machine learning has brought renewed motivation for smooth games. In this work, we focus on the c… (voir plus)lass of stochastic Hamiltonian methods and provide the first convergence guarantees for certain classes of stochastic smooth games. We propose a novel unbiased estimator for the stochastic Hamiltonian gradient descent (SHGD) and highlight its benefits. Using tools from the optimization literature we show that SHGD converges linearly to the neighbourhood of a stationary point. To guarantee convergence to the exact solution, we analyze SHGD with a decreasing step-size and we also present the first stochastic variance reduced Hamiltonian method. Our results provide the first global non-asymptotic last-iterate convergence guarantees for the class of stochastic unconstrained bilinear games and for the more general class of stochastic games that satisfy a "sufficiently bilinear" condition, notably including some non-convex non-concave problems. We supplement our analysis with experiments on stochastic bilinear and sufficiently bilinear games, where our theory is shown to be tight, and on simple adversarial machine learning formulations.
Drawing an inspiration from behavioral studies of human decision making, we propose here a more general and flexible parametric framework fo… (voir plus)r reinforcement learning that extends standard Q-learning to a two-stream model for processing positive and negative rewards, and allows to incorporate a wide range of reward-processing biases -- an important component of human decision making which can help us better understand a wide spectrum of multi-agent interactions in complex real-world socioeconomic systems, as well as various neuropsychiatric conditions associated with disruptions in normal reward processing. From the computational perspective, we observe that the proposed Split-QL model and its clinically inspired variants consistently outperform standard Q-Learning and SARSA methods, as well as recently proposed Double Q-Learning approaches, on simulated tasks with particular reward distributions, a real-world dataset capturing human decision-making in gambling tasks, and the Pac-Man game in a lifelong learning setting across different reward stationarities.
Structured Conditional Continuous Normalizing Flows for Efficient Amortized Inference in Graphical Models
Christian Dietrich Weilbach
Boyan Beronov
F. Wood
William Harvey
We exploit minimally faithful inversion of graphical model structures to specify sparse continuous normalizing flows (CNFs) for amortized i… (voir plus)nference. We find that the sparsity of this factorization can be exploited to reduce the numbers of parameters in the neural network, adaptive integration steps of the flow, and consequently FLOPs at both training and inference time without decreasing performance in comparison to unconstrained flows. By expressing the structure inversion as a compilation pass in a probabilistic programming language, we are able to apply it in a novel way to models as complex as convolutional neural networks. Furthermore, we extend the training objective for CNFs in the context of inference amortization to the symmetric Kullback-Leibler divergence, and demonstrate its theoretical and practical advantages.
2020-01-01
International Conference on Artificial Intelligence and Statistics (publié)
Structured Conditional Continuous Normalizing Flows for Efficient Amortized Inference in Graphical Models
Christian Dietrich Weilbach
Boyan Beronov
F. Wood
William Harvey
We exploit minimally faithful inversion of graphical model structures to specify sparse continuous normalizing flows (CNFs) for amortized i… (voir plus)nference. We find that the sparsity of this factorization can be exploited to reduce the numbers of parameters in the neural network, adaptive integration steps of the flow, and consequently FLOPs at both training and inference time without decreasing performance in comparison to unconstrained flows. By expressing the structure inversion as a compilation pass in a probabilistic programming language, we are able to apply it in a novel way to models as complex as convolutional neural networks. Furthermore, we extend the training objective for CNFs in the context of inference amortization to the symmetric Kullback-Leibler divergence, and demonstrate its theoretical and practical advantages.
2020-01-01
International Conference on Artificial Intelligence and Statistics (published)