Publications

Separating value functions across time-scales
Joshua Romoff
Peter Henderson
Ahmed Touati
Emma Brunskill
Joelle Pineau
Yann Ollivier
In many finite horizon episodic reinforcement learning (RL) settings, it is desirable to optimize for the undiscounted return - in settings … (see more)like Atari, for instance, the goal is to collect the most points while staying alive in the long run. Yet, it may be difficult (or even intractable) mathematically to learn with this target. As such, temporal discounting is often applied to optimize over a shorter effective planning horizon. This comes at the risk of potentially biasing the optimization target away from the undiscounted goal. In settings where this bias is unacceptable - where the system must optimize for longer horizons at higher discounts - the target of the value function approximator may increase in variance leading to difficulties in learning. We present an extension of temporal difference (TD) learning, which we call TD(
2019-05-23
International Conference on Machine Learning (published)
doi.org
proceedings.mlr.press
State-Reification Networks: Improving Generalization by Modeling the Distribution of Hidden Representations
Alex Lamb
Jonathan Binas
Anirudh Goyal
Sandeep Subramanian
Ioannis Mitliagkas
Denis Kazakov
Yoshua Bengio
Michael C. Mozer
Machine learning promises methods that generalize well from finite labeled data. However, the brittleness of existing neural net approaches … (see more)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.
2019-05-23
International Conference on Machine Learning (unknown)
doi.org
proceedings.mlr.press
Stroke Lesion Segmentation in FLAIR MRI Datasets Using Customized Markov Random Fields
Nagesh K. Subbanna
Deepthi Rajashekar
Bastian Cheng
Götz Thomalla
Jens Fiehler
Tal Arbel
Nils D. Forkert
Robust and reliable stroke lesion segmentation is a crucial step toward employing lesion volume as an independent endpoint for randomized tr… (see more)ials. The aim of this work was to develop and evaluate a novel method to segment sub-acute ischemic stroke lesions from fluid-attenuated inversion recovery (FLAIR) magnetic resonance imaging (MRI) datasets. After preprocessing of the datasets, a Bayesian technique based on Gabor textures extracted from the FLAIR signal intensities is utilized to generate a first estimate of the lesion segmentation. Using this initial segmentation, a customized voxel-level Markov random field model based on intensity as well as Gabor texture features is employed to refine the stroke lesion segmentation. The proposed method was developed and evaluated based on 151 multi-center datasets from three different databases using a leave-one-patient-out validation approach. The comparison of the automatically segmented stroke lesions with manual ground truth segmentation revealed an average Dice coefficient of 0.582, which is in the upper range of previously presented lesion segmentation methods using multi-modal MRI datasets. Furthermore, the results obtained by the proposed technique are superior compared to the results obtained by two methods based on convolutional neural networks and three phase level-sets, respectively, which performed best in the ISLES 2015 challenge using multi-modal imaging datasets. The results of the quantitative evaluation suggest that the proposed method leads to robust lesion segmentation results using FLAIR MRI datasets only as a follow-up sequence.
2019-05-23
Frontiers in Neurology (published)
doi.org
The Value Function Polytope in Reinforcement Learning
Robert Dadashi
Adrien Ali Taiga
Nicolas Roux
Dale Schuurmans
Bellemare Marc-Emmanuel
We establish geometric and topological properties of the space of value functions in finite state-action Markov decision processes. Our main… (see more) contribution is the characterization of the nature of its shape: a general polytope (Aigner et al., 2010). To demonstrate this result, we exhibit several properties of the structural relationship between policies and value functions including the line theorem, which shows that the value functions of policies constrained on all but one state describe a line segment. Finally, we use this novel perspective to introduce visualizations to enhance the understanding of the dynamics of reinforcement learning algorithms.
2019-05-23
Proceedings of the 36th International Conference on Machine Learning (published)
proceedings.mlr.press
proceedings.mlr.press
Understanding the impact of entropy on policy optimization
Zafarali Ahmed
Nicolas Roux
Mohammad Norouzi
Dale Schuurmans
Entropy regularization is commonly used to improve policy optimization in reinforcement learning. It is believed to help with \emph{explorat… (see more)ion} by encouraging the selection of more stochastic policies. In this work, we analyze this claim using new visualizations of the optimization landscape based on randomly perturbing the loss function. We first show that even with access to the exact gradient, policy optimization is difficult due to the geometry of the objective function. Then, we qualitatively show that in some environments, a policy with higher entropy can make the optimization landscape smoother, thereby connecting local optima and enabling the use of larger learning rates. This paper presents new tools for understanding the optimization landscape, shows that policy entropy serves as a regularizer, and highlights the challenge of designing general-purpose policy optimization algorithms.
2019-05-23
Proceedings of the 36th International Conference on Machine Learning (published)
proceedings.mlr.press
proceedings.mlr.press
Recurrent Value Functions
Pierre Thodoroff
Nishanth Anand
Lucas Caccia
Doina Precup
Joelle Pineau
Despite recent successes in Reinforcement Learning, value-based methods often suffer from high variance hindering performance. In this paper… (see more), we illustrate this in a continuous control setting where state of the art methods perform poorly whenever sensor noise is introduced. To overcome this issue, we introduce Recurrent Value Functions (RVFs) as an alternative to estimate the value function of a state. We propose to estimate the value function of the current state using the value function of past states visited along the trajectory. Due to the nature of their formulation, RVFs have a natural way of learning an emphasis function that selectively emphasizes important states. First, we establish RVF's asymptotic convergence properties in tabular settings. We then demonstrate their robustness on a partially observable domain and continuous control tasks. Finally, we provide a qualitative interpretation of the learned emphasis function.
2019-05-22
arXiv (Cornell University) (preprint)
doi.org
arxiv.org
The Journey is the Reward: Unsupervised Learning of Influential Trajectories
Jonathan Binas
Sherjil Ozair
Yoshua Bengio
Unsupervised exploration and representation learning become increasingly important when learning in diverse and sparse environments. The inf… (see more)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.
2019-05-21
ArXiv (preprint)
arxiv.org
arxiv.org
A Data-Efficient Framework for Training and Sim-to-Real Transfer of Navigation Policies
Homanga Bharadhwaj
Zihan Wang
Yoshua Bengio
Liam Paull
Learning effective visuomotor policies for robots purely from data is challenging, but also appealing since a learning-based system should n… (see more)ot require manual tuning or calibration. In the case of a robot operating in a real environment the training process can be costly, time-consuming, and even dangerous since failures are common at the start of training. For this reason, it is desirable to be able to leverage \textit{simulation} and \textit{off-policy} data to the extent possible to train the robot. In this work, we introduce a robust framework that plans in simulation and transfers well to the real environment. Our model incorporates a gradient-descent based planning module, which, given the initial image and goal image, encodes the images to a lower dimensional latent state and plans a trajectory to reach the goal. The model, consisting of the encoder and planner modules, is trained through a meta-learning strategy in simulation first. We subsequently perform adversarial domain transfer on the encoder by using a bank of unlabelled but random images from the simulation and real environments to enable the encoder to map images from the real and simulated environments to a similarly distributed latent representation. By fine tuning the entire model (encoder + planner) with far fewer real world expert demonstrations, we show successful planning performances in different navigation tasks.
2019-05-19
2019 International Conference on Robotics and Automation (ICRA) (published)
doi.org
arxiv.org
Semantic Mapping for View-Invariant Relocalization.
Jimmy Li
David Meger
Gregory Dudek
We propose a system for visual simultaneous localization and mapping (SLAM) that combines traditional local appearance-based features with s… (see more)emantically meaningful object landmarks to achieve both accurate local tracking and highly view-invariant object-driven relocalization. Our mapping process uses a sampling-based approach to efficiently infer the 3D pose of object landmarks from 2D bounding box object detections. These 3D landmarks then serve as a view-invariant representation which we leverage to achieve camera relocalization even when the viewing angle changes by more than 125 degrees. This level of view-invariance cannot be attained by local appearance-based features (e.g. SIFT) since the same set of surfaces are not even visible when the viewpoint changes significantly. Our experiments show that even when existing methods fail completely for viewpoint changes of more than 70 degrees, our method continues to achieve a relocalization rate of around 90%, with a mean rotational error of around 8 degrees.
2019-05-19
2019 International Conference on Robotics and Automation (ICRA) (published)
doi.org
Uncertainty Aware Learning from Demonstrations in Multiple Contexts using Bayesian Neural Networks
Sanjay Thakur
Herke van Hoof
Juan Camilo Gamboa Higuera
Doina Precup
David Meger
Diversity of environments is a key challenge that causes learned robotic controllers to fail due to the discrepancies between the training a… (see more)nd evaluation conditions. Training from demonstrations in various conditions can mitigate---but not completely prevent---such failures. Learned controllers such as neural networks typically do not have a notion of uncertainty that allows to diagnose an offset between training and testing conditions, and potentially intervene. In this work, we propose to use Bayesian Neural Networks, which have such a notion of uncertainty. We show that uncertainty can be leveraged to consistently detect situations in high-dimensional simulated and real robotic domains in which the performance of the learned controller would be sub-par. Also, we show that such an uncertainty based solution allows making an informed decision about when to invoke a fallback strategy. One fallback strategy is to request more data. We empirically show that providing data only when requested results in increased data-efficiency.
2019-05-19
2019 International Conference on Robotics and Automation (ICRA) (published)
doi.org
arxiv.org
A Highly Adaptive Acoustic Model for Accurate Multi-dialect Speech Recognition
Sanghyun Yoo
Inchul Song
Yoshua Bengio
Despite the success of deep learning in speech recognition, multi-dialect speech recognition remains a difficult problem. Although dialect-s… (see more)pecific acoustic models are known to perform well in general, they are not easy to maintain when dialect-specific data is scarce and the number of dialects for each language is large. Therefore, a single unified acoustic model (AM) that generalizes well for many dialects has been in demand. In this paper, we propose a novel acoustic modeling technique for accurate multi-dialect speech recognition with a single AM. Our proposed AM is dynamically adapted based on both dialect information and its internal representation, which results in a highly adaptive AM for handling multiple dialects simultaneously. We also propose a simple but effective training method to deal with unseen dialects. The experimental results on large scale speech datasets show that the proposed AM outperforms all the previous ones, reducing word error rates (WERs) by 8.11% relative compared to a single all-dialects AM and by 7.31% relative compared to dialect-specific AMs.
2019-05-11
ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (published)
doi.org
arxiv.org
Representation Mixing for TTS Synthesis
Kyle Kastner
Joao Felipe Santos
Yoshua Bengio
Aaron Courville
Recent character and phoneme-based parametric TTS systems using deep learning have shown strong performance in natural speech generation. Ho… (see more)wever, the choice between character or phoneme input can create serious limitations for practical deployment, as direct control of pronunciation is crucial in certain cases. We demonstrate a simple method for combining multiple types of linguistic information in a single encoder, named representation mixing, enabling flexible choice between character, phoneme, or mixed representations during inference. Experiments and user studies on a public audiobook corpus show the efficacy of our approach.
2019-05-11
ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (published)
doi.org
arxiv.org