Dumitru Erhan

Dumitru Erhan
Google Inc. | Google · Research Department

PhD

About

55
Publications
160,963
Reads
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
90,610
Citations
Additional affiliations
September 2004 - February 2011
Université de Montréal
Position
  • PhD Student

Publications

Publications (55)
Preprint
Full-text available
Model-based reinforcement learning (RL) algorithms designed for handling complex visual observations typically learn some sort of latent state representation, either explicitly or implicitly. Standard methods of this sort do not distinguish between functionally relevant aspects of the state and irrelevant distractors, instead aiming to represent al...
Preprint
Full-text available
An agent that is capable of predicting what happens next can perform a variety of tasks through planning with no additional training. Furthermore, such an agent can internally represent the complex dynamics of the real-world and therefore can acquire a representation useful for a variety of visual perception tasks. This makes predicting the future...
Preprint
Full-text available
Model-based reinforcement learning (MBRL) methods have shown strong sample efficiency and performance across a variety of tasks, including when faced with high-dimensional visual observations. These methods learn to predict the environment dynamics and expected reward from interaction and use this predictive model to plan and perform the task. Howe...
Preprint
Full-text available
Autonomous driving system development is critically dependent on the ability to replay complex and diverse traffic scenarios in simulation. In such scenarios, the ability to accurately simulate the vehicle sensors such as cameras, lidar or radar is essential. However, current sensor simulators leverage gaming engines such as Unreal or Unity, requir...
Conference Paper
Full-text available
We propose an empirical measure of the approximate accuracy of feature importance estimates in deep neural networks. Our results across several large-scale image classification datasets show that many popular interpretability methods produce estimates of feature importance that are not better than a random designation of feature importance. Only ce...
Preprint
Full-text available
Predicting future video frames is extremely challenging, as there are many factors of variation that make up the dynamics of how frames change through time. Previously proposed solutions require complex inductive biases inside network architectures with highly specialized computation, including segmentation masks, optical flow, and foreground and b...
Chapter
Full-text available
Saliency methods aim to explain the predictions of deep neural networks. These methods lack reliability when the explanation is sensitive to factors that do not contribute to the model prediction. We use a simple and common pre-processing step which can be compensated for easily—adding a constant shift to the input data—to show that a transformatio...
Preprint
Full-text available
Generative models that can model and predict sequences of future events can, in principle, learn to capture complex real-world phenomena, such as physical interactions. In particular, learning predictive models of videos offers an especially appealing mechanism to enable a rich understanding of the physical world: videos of real-world interactions...
Preprint
Full-text available
Model-free reinforcement learning (RL) can be used to learn effective policies for complex tasks, such as Atari games, even from image observations. However, this typically requires very large amounts of interaction -- substantially more, in fact, than a human would need to learn the same games. How can people learn so quickly? Part of the answer m...
Preprint
Full-text available
Estimating the influence of a given feature to a model prediction is challenging. We introduce ROAR, RemOve And Retrain, a benchmark to evaluate the accuracy of interpretability methods that estimate input feature importance in deep neural networks. We remove a fraction of input features deemed to be most important according to each estimator and m...
Preprint
Full-text available
Much of recent research has been devoted to video prediction and generation, yet most of the previous works have demonstrated only limited success in generating videos on short-term horizons. The hierarchical video prediction method by Villegas et al. (2017) is an example of a state-of-the-art method for long-term video prediction, but their method...
Article
Full-text available
Saliency methods aim to explain the predictions of deep neural networks. These methods lack reliability when the explanation is sensitive to factors that do not contribute to the model prediction. We use a simple and common pre-processing step ---adding a constant shift to the input data--- to show that a transformation with no effect on the model...
Article
Full-text available
Predicting the future in real-world settings, particularly from raw sensory observations such as images, is exceptionally challenging. Real-world events can be stochastic and unpredictable, and the high dimensionality and complexity of natural images requires the predictive model to build an intricate understanding of the natural world. Many existi...
Article
Full-text available
DeConvNet, Guided BackProp, LRP, were invented to better understand deep neural networks. We show that these methods do not produce the theoretically correct explanation for a linear model. Yet they are used on multi-layer networks with millions of parameters. This is a cause for concern since linear models are simple neural networks. We argue that...
Article
Full-text available
Collecting well-annotated image datasets to train modern machine learning algorithms is prohibitively expensive for many tasks. One appealing alternative is rendering synthetic data where ground-truth annotations are generated automatically. Unfortunately, models trained purely on rendered images often fail to generalize to real images. To address...
Conference Paper
Full-text available
We present a method for detecting objects in images using a single deep neural network. Our approach, named SSD, discretizes the output space of bounding boxes into a set of default boxes over different aspect ratios and scales per feature map location. At prediction time, the network generates scores for the presence of each object category in eac...
Article
Full-text available
The cost of large scale data collection and annotation often makes the application of machine learning algorithms to new tasks or datasets prohibitively expensive. One approach circumventing this cost is training models on synthetic data where annotations are provided automatically. Despite their appeal, such models often fail to generalize from sy...
Article
Full-text available
Automatically describing the content of an image is a fundamental problem in artificial intelligence that connects computer vision and natural language processing. In this paper, we present a generative model based on a deep recurrent architecture that combines recent advances in computer vision and machine translation and that can be used to gener...
Article
Full-text available
Theano is a Python library that allows to define, optimize, and evaluate mathematical expressions involving multi-dimensional arrays efficiently. Since its introduction, it has been one of the most used CPU and GPU mathematical compilers - especially in the machine learning community - and has shown steady performance improvements. Theano is being...
Article
Full-text available
We present a method for detecting objects in images using a single deep neural network. Our approach, named SSD, discretizes the output space of bounding boxes into a set of bounding box priors over different aspect ratios and scales per feature map location. At prediction time, the network generates confidences that each prior corresponds to objec...
Article
Full-text available
Current state-of-the-art deep learning systems for visual object recognition and detection use purely supervised training with regularization such as dropout to avoid overfitting. The performance depends critically on the amount of labeled examples, and in current practice the labels are assumed to be unambiguous and accurate. However, this assumpt...
Article
Full-text available
Most high quality object detection approaches use the same scheme: salience-based object proposal methods followed by post-classification using deep convolutional features. In this work, we demonstrate that fully learnt, data-driven proposal generation methods can effectively match the accuracy of their hand engineered counterparts, while allowing...
Article
Full-text available
We study sequential decision making in environments where rewards are only partially observed, but can be modeled as a function of observed contexts and the chosen action by the decision maker. This setting, known as contextual bandits, encompasses a wide variety of applications such as health care, content recommendation and Internet advertising....
Article
Full-text available
Automatically describing the content of an image is a fundamental problem in artificial intelligence that connects computer vision and natural language processing. In this paper, we present a generative model based on a deep recurrent architecture that combines recent advances in computer vision and machine translation and that can be used to gener...
Article
Full-text available
We propose a deep convolutional neural network architecture codenamed "Inception", which was responsible for setting the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC 2014). The main hallmark of this architecture is the improved utilization of the computing resources insi...
Article
Full-text available
Deep neural networks are highly expressive models that have recently achieved state of the art performance on speech and visual recognition tasks. While their expressiveness is the reason they succeed, it also causes them to learn uninterpretable solutions that could have counter-intuitive properties. In this paper we report two such properties. Fi...
Article
Full-text available
Object recognition and localization are important tasks in computer vision. The focus of this work is the incorporation of contextual information in order to improve object recognition and localization. For instance, it is natural to expect not to see an elephant to appear in the middle of an ocean. We consider a simple approach to encapsulate such...
Article
Full-text available
Deep convolutional neural networks have recently achieved state-of-the-art performance on a number of image recognition benchmarks, including the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC-2012). The winning model on the localization sub-task was a network that predicts a single bounding box and a confidence score for each object cat...
Article
Full-text available
The ICML 2013 Workshop on Challenges in Representation Learning. 11http://deeplearning.net/icml2013-workshop-competition. focused on three challenges: the black box learning challenge, the facial expression recognition challenge, and the multimodal learning challenge. We describe the datasets created for these challenges and summarize the results o...
Article
Full-text available
Deep Neural Networks (DNNs) have recently shown outstanding performance on image classification tasks [14]. In this paper we go one step further and address the problem of object detection using DNNs, that is not only classifying but also precisely localizing objects of various classes. We present a simple and yet powerful formulation of object det...
Conference Paper
Full-text available
Deep Neural Networks (DNNs) have recently shown outstanding performance on image classification tasks [14]. In this paper we go one step further and address the problem of object detection using DNNs, that is not only classifying but also precisely localizing objects of various classes. We present a simple and yet powerful formulation of object det...
Article
Full-text available
We present and prove properties of a new offline policy evaluator for an exploration learning setting which is superior to previous evaluators. In particular, it simultaneously and correctly incorporates techniques from importance weighting, doubly robust evaluation, and nonstationary policy evaluation approaches. In addition, our approach allows g...
Article
Full-text available
Recent theoretical and empirical work in statistical machine learning has demonstrated the potential of learning algorithms for deep architectures, i.e., function classes obtained by composing multiple levels of representation. The hypothesis evaluated here is that intermediate levels of representation, because they can be shared across tasks and e...
Article
Full-text available
Recent theoretical and empirical work in statistical machine learning has demonstrated the importance of learning algorithms for deep architectures, i.e., function classes obtained by composing multiple non-linear transformations. Self-taught learning (exploiting unlabeled examples or examples from other distributions) has already been applied to d...
Article
Full-text available
Much recent research has been devoted to learning algorithms for deep architectures such as Deep Belief Networks and stacks of auto-encoder variants, with impressive results obtained in several areas, mostly on vision and language data sets. The best results obtained on supervised learning tasks involve an unsupervised learning component, usually i...
Article
Full-text available
Much recent research has been devoted to learning algorithms for deep architectures such as Deep Belief Networks and stacks of auto-encoder variants, with impressive results obtained in several areas, mostly on vision and language data sets. The best results obtained on supervised learning tasks involve an unsupervised learning component, usually i...
Article
Full-text available
We investigate a simple yet effective method to introduce inhibitory and excitatory interactions between units in the layers of a deep neural network classifier. The method is based on the greedy layer-wise procedure of deep learning algorithms and extends the denoising autoencoder (Vincent et al., 2008) by adding asymmetric lateral connections bet...
Article
Full-text available
Whereas theoretical work suggests that deep ar- chitectures might be more efficient at represent- ing highly-varying functions, training deep ar- chitectures was unsuccessful until the recent ad- vent of algorithms based on unsupervised pre- training. Even though these new algorithms have enabled training deep models, many questions remain as to th...
Article
Full-text available
Deep architectures have demonstrated state-of-the-art results in a variety of settings, especially with vision datasets. Beyond the model definitions and the quantitative analyses, there is a need for qualitative comparisons of the solutions learned by various deep architectures. The goal of this paper is to find good qualita-tive interpretations o...
Conference Paper
Full-text available
We introduce the problem of zero-data learning, where a model must generalize to classes or tasks for which no train- ing data are available and only a description of the classes or tasks are provided. Zero-data learning is useful for prob- lems where the set of classes to distinguish or tasks to solve is very large and is not entirely covered by t...
Conference Paper
Full-text available
Recently, several learning algorithms rely- ing on models with deep architectures have been proposed. Though they have demon- strated impressive performance, to date, they have only been evaluated on relatively simple problems such as digit recognition in a con- trolled environment, for which many machine learning algorithms already report reasonab...
Article
Full-text available
We present an algorithm that predicts musical genre and artist from an audio waveform. Our method uses the ensemble learner AdaBoost to select from a set of audio features that have been extracted from segmented audio and then aggregated. Our classifier proved to be the most eective method for genre classification at the recent MIREX 2005 internati...
Article
Full-text available
We investigate the problem of learning several tasks simultaneously in order to transfer the acquired knowledge to a completely new task for which no training data are available. Assuming that the tasks share some representation that we can discover efficiently, such a scenario should lead to a better model of the new task, as compared to the model...
Article
Full-text available
Building a QSAR model of a new biological target for which few screening data are available is a statistical challenge. However, the new target may be part of a bigger family, for which we have more screening data. Collaborative filtering or, more generally, multi-task learning, is a machine learning approach that improves the generalization perfor...
Article
Full-text available
This paper reviews the Maximum Likelihood estimation problem and its solution via the Expectation-Maximization algorithm. Emphasis is made on the description of finite mixtures of multi-variate Bernoulli distributions for modeling 0-1 data. General ideas about convergence and non-identifiability are presented. We discuss improvements to the algorit...
Article
Full-text available
Deep architectures have demonstrated state-of-the-art performance in a variety of settings, especially with vision datasets. Deep learning algorithms are based on learn-ing several levels of representation of the input. Beyond test-set performance, there is a need for qualitative comparisons of the solutions learned by various deep archi-tectures,...
Article
Full-text available
We investigate the use of Hessian Free optimization for learning deep au-toencoders. One of the critical components in that algorithm is the choice of the preconditioner. We argue in this paper that the Jacobi precondi-tioner leads to faster optimization and we show how it can be accurately and efficiently estimated using a randomized algorithm.

Network

Cited By

Projects

Project (1)
Project
Deep learning has significantly advanced the state of the art in machine learning. However, neural networks are often considered black boxes. There is significant effort to develop techniques that explain a classifier’s decisions. Although some of these approaches have resulted in compelling visualisations, there is a lack of theory of what is actually explained. Here we present an analysis of these methods and formulate a quality criterion for explanation methods. On this ground, we propose methods that may serve as an extension for existing back-projection and decomposition techniques.