Jonathan Binas

Charles Blundell

Michael Curtis Mozer

Deep learning has seen a movement away from representing examples with a monolithic hidden state towards a richly structured state. For exam… (voir plus)ple, Transformers segment by position, and object-centric architectures decompose images into entities. In all these architectures, interactions between different elements are modeled via pairwise interactions: Transformers make use of self-attention to incorporate information from other positions and object-centric architectures make use of graph neural networks to model interactions among entities. We consider how to improve on pairwise interactions in terms of global coordination and a coherent, integrated representation that can be used for downstream tasks. In cognitive science, a global workspace architecture has been proposed in which functionally specialized components share information through a common, bandwidth-limited communication channel. We explore the use of such a communication channel in the context of deep learning for modeling the structure of complex environments. The proposed method includes a shared workspace through which communication among different specialist modules takes place but due to limits on the communication bandwidth, specialist modules must compete for access. We show that capacity limitations have a rational basis in that (1) they encourage specialization and compositionality and (2) they facilitate the synchronization of otherwise independent specialists.

2022-01-28

ICLR.cc/2022/Conference (présentation orale)

Reinforcement Learning with Random Delays

Simon Ramstedt

Action and observation delays commonly occur in many Reinforcement Learning applications, such as remote control scenarios. We study the ana… (voir plus)tomy of randomly delayed environments, and show that partially resampling trajectory fragments in hindsight allows for off-policy multi-step value estimation. We apply this principle to derive Delay-Correcting Actor-Critic (DCAC), an algorithm based on Soft Actor-Critic with significantly better performance in environments with delays. This is shown theoretically and also demonstrated practically on a delay-augmented version of the MuJoCo continuous control benchmark.

2021-01-12

ICLR.cc/2021/Conference (poster)

Out-of-Distribution Generalization via Risk Extrapolation (REx)

David Scott Krueger

Ethan Caballero

Joern-Henrik Jacobsen

Amy Zhang

Rémi LE PRIOL

Aaron Courville

Generalizing outside of the training distribution is an open challenge for current machine learning systems. A weak form of out-of-distribut… (voir plus)ion (OoD) generalization is the ability to successfully interpolate between multiple observed distributions. One way to achieve this is through robust optimization, which seeks to minimize the worst-case risk over convex combinations of the training distributions. However, a much stronger form of OoD generalization is the ability of models to extrapolate beyond the distributions observed during training. In pursuit of strong OoD generalization, we introduce the principle of Risk Extrapolation (REx). REx can be viewed as encouraging robustness over affine combinations of training risks, by encouraging strict equality between training risks. We show conceptually how this principle enables extrapolation, and demonstrate the effectiveness and scalability of instantiations of REx on various OoD generalization tasks. Our code can be found at this https URL.

2020-03-02

ArXiv (prépublication)

Reinforcement Learning with Competitive Ensembles of Information-Constrained Primitives

Anirudh Goyal

Shagun Sodhani

Xue Bin Peng

Sergey Levine

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.

2020-01-01

ICLR.cc/2020/Conference (poster)

Retrieving Signals with Deep Complex Extractors

Ousmane Dia

Recent advances have made it possible to create deep complex-valued neural networks. Despite this progress, many challenging learning tasks … (voir plus)have yet to leverage the power of complex representations. Building on recent advances, we propose a new deep complex-valued method for signal retrieval and extraction in the frequency domain. As a case study, we perform audio source separation in the Fourier domain. Our new method takes advantage of the convolution theorem which states that the Fourier transform of two convolved signals is the elementwise product of their Fourier transforms. Our novel method is based on a complex-valued version of Feature-Wise Linear Modulation (FiLM) and serves as the keystone of our proposed signal extraction method. We also introduce a new and explicit amplitude and phase-aware loss, which is scale and time invariant, taking into account the complex-valued components of the spectrogram. Using the Wall Street Journal Dataset, we compared our phase-aware loss to several others that operate both in the time and frequency domains and demonstrate the effectiveness of our proposed signal extraction method and proposed loss.

2019-10-21

NeurIPS.cc/2019/Workshop/Deep_Inverse (poster)

State-Reification Networks: Improving Generalization by Modeling the Distribution of Hidden Representations

Denis Kazakov

Michael Curtis Mozer

Machine learning promises methods that generalize well from finite labeled data. However, the brittleness of existing neural net approaches … (voir plus)is revealed by notable failures, such as the existence of adversarial examples that are misclassified despite being nearly identical to a training example, or the inability of recurrent sequence-processing nets to stay on track without teacher forcing. We introduce a method, which we refer to as \emph{state reification}, that involves modeling the distribution of hidden states over the training data and then projecting hidden states observed during testing toward this distribution. Our intuition is that if the network can remain in a familiar manifold of hidden space, subsequent layers of the net should be well trained to respond appropriately. We show that this state-reification method helps neural nets to generalize better, especially when labeled data are sparse, and also helps overcome the challenge of achieving robust generalization with adversarial training.

2019-05-24

Proceedings of the 36th International Conference on Machine Learning (publié)

proceedings.mlr.press

The Journey is the Reward: Unsupervised Learning of Influential Trajectories

Sherjil Ozair

Unsupervised exploration and representation learning become increasingly important when learning in diverse and sparse environments. The inf… (voir plus)ormation-theoretic principle of empowerment formalizes an unsupervised exploration objective through an agent trying to maximize its influence on the future states of its environment. Previous approaches carry certain limitations in that they either do not employ closed-loop feedback or do not have an internal state. As a consequence, a privileged final state is taken as an influence measure, rather than the full trajectory. We provide a model-free method which takes into account the whole trajectory while still offering the benefits of option-based approaches. We successfully apply our approach to settings with large action spaces, where discovery of meaningful action sequences is particularly difficult.

2019-05-22

ArXiv (prépublication)

Lagrangian neurodynamics for real-time error-backpropagation across cortical areas

Dominik Dold

Akos F. Kungl

João Sacramento

Mihai A. Petrovici

Kaspar Anton Schindler

Walter Senn

.

Reinforcement Learning for Sustainable Agriculture

Leonie H. Luginbuehl

Modern machine learning methods have achieved superhuman performance on a variety of tasks, simply learning from the outcomes of their actio… (voir plus)ns. We propose a path towards more sustainable agriculture, considering plant development an optimization problem with respect to certain parameters, such as yield and environmental impact, which can be optimized in an automated way. Specifically, we propose to use reinforcement learning to autonomously explore and learn ways of influencing the development of certain types of plants, controlling environmental parameters, such as irrigation or nutrient supply, and receiving sensory feedback, such as camera images, humidity, and moisture measurements. The trained system will thus be able to provide instructions for optimal treatment of a local population of plants, based on non-invasive measurements, such as imaging.

Generalization of Equilibrium Propagation to Vector Field Dynamics

The biological plausibility of the backpropagation algorithm has long been doubted by neuroscientists. Two major reasons are that neurons wo… (voir plus)uld need to send two different types of signal in the forward and backward phases, and that pairs of neurons would need to communicate through symmetric bidirectional connections. We present a simple two-phase learning procedure for fixed point recurrent networks that addresses both these issues. In our model, neurons perform leaky integration and synaptic weights are updated through a local mechanism. Our learning method generalizes Equilibrium Propagation to vector field dynamics, relaxing the requirement of an energy function. As a consequence of this generalization, the algorithm does not compute the true gradient of the objective function, but rather approximates it at a precision which is proven to be directly related to the degree of symmetry of the feedforward and feedback weights. We show experimentally that our algorithm optimizes the objective function.

2018-08-14

ArXiv (prépublication)

Low-memory convolutional neural networks through incremental depth-first processing

We introduce an incremental processing scheme for convolutional neural network (CNN) inference, targeted at embedded applications with limit… (voir plus)ed memory budgets. Instead of processing layers one by one, individual input pixels are propagated through all parts of the network they can influence under the given structural constraints. This depth-first updating scheme comes with hard bounds on the memory footprint: the memory required is constant in the case of 1D input and proportional to the square root of the input dimension in the case of 2D input.

2018-04-28

ArXiv (prépublication)