35 results found
Wang R, Ciliberto C, Amadori P, et al., 2019, Random Expert Distillation: Imitation Learning via Expert Policy Support Estimation, Thirty-sixth International Conference on Machine Learning, Publisher: Proceedings of International Conference on Machine Learning (ICML-2019)
We consider the problem of imitation learning from a finite set of experttrajectories, without access to reinforcement signals. The classical approachof extracting the expert's reward function via inverse reinforcement learning,followed by reinforcement learning is indirect and may be computationallyexpensive. Recent generative adversarial methods based on matching the policydistribution between the expert and the agent could be unstable duringtraining. We propose a new framework for imitation learning by estimating thesupport of the expert policy to compute a fixed reward function, which allowsus to re-frame imitation learning within the standard reinforcement learningsetting. We demonstrate the efficacy of our reward function on both discreteand continuous domains, achieving comparable or better performance than thestate of the art under different reinforcement learning algorithms.
Pasquale G, Ciliberto C, Odone F, et al., 2019, Are we done with object recognition? The iCub robot's perspective, ROBOTICS AND AUTONOMOUS SYSTEMS, Vol: 112, Pages: 260-281, ISSN: 0921-8890
Luise G, Stamos D, Pontil M, et al., 2019, Leveraging low-rank relations between surrogate tasks in structured prediction, 36th International Conference on Machine Learning, ICML 2019, Vol: 2019-June, Pages: 7415-7444
© 36th International Conference on Machine Learning, ICML 2019. All rights reserved. We study the interplay between surrogate methods for structured prediction and techniques from multitask learning designed to leverage relationships between surrogate outputs. We propose an efficient algorithm based on trace norm regularization which, differently from previous methods, does not require explicit knowledge of the coding/decoding functions of the surrogate framework. As a result, our algorithm can be applied to the broad class of problems in which the surrogate space is large or even infinite dimensional. We study excess risk bounds for trace norm regularized structured prediction proving the consistency and learning rates for our estimator. We also identify relevant regimes in which our approach can enjoy better generalization performance than previous methods. Numerical experiments on ranking problems indicate that enforcing low-rank relations among surrogate outputs may indeed provide a significant advantage in practice.
Denevi G, Ciliberto C, Grazzi R, et al., 2019, Learning-to-learn stochastic gradient descent with biased regularization
© 36th International Conference on Machine Learning, ICML 2019. All rights reserved. We study the problem of learning-to-learn: inferring a learning algorithm that works well on a family of tasks sampled from an unknown distribution. As class of algorithms we consider Stochastic Gradient Descent (SGD) on the true risk regularized by the square euclidean distance from a bias vector. We present an average excess risk bound for such a learning algorithm that quantifies the potential benefit of using a bias vector with respect to the unbiased case. We then propose a novel meta-algorithm to estimate the bias term online from a sequence of observed tasks. The small memory footprint and low time complexity of our approach makes it appealing in practice while our theoretical analysis provides guarantees on the generalization properties of the meta-algorithm on new tasks. A key feature of our results is that, when the number of tasks grows and their variance is relatively small, our learning-to-learn approach has a significant advantage over learning each task in isolation by standard SGD without a bias term. Numerical experiments demonstrate the effectiveness of our approach in practice.
Luise G, Rudi A, Pontil M, et al., 2018, Differential Properties of Sinkhorn Approximation for Learning with Wasserstein Distance, Advances in Neural Information Processing Systems (NeurIPS), Dec 2018, Montr\'eal, Canada
Applications of optimal transport have recently gained remarkable attentionthanks to the computational advantages of entropic regularization. However, inmost situations the Sinkhorn approximation of the Wasserstein distance isreplaced by a regularized version that is less accurate but easy todifferentiate. In this work we characterize the differential properties of theoriginal Sinkhorn distance, proving that it enjoys the same smoothness as itsregularized version and we explicitly provide an efficient algorithm to computeits gradient. We show that this result benefits both theory and applications:on one hand, high order smoothness confers statistical guarantees to learningwith Wasserstein approximations. On the other hand, the gradient formula allowsus to efficiently solve learning and optimization problems in practice.Promising preliminary experiments complement our analysis.
Rudi A, Ciliberto C, Marconi GM, et al., 2018, Manifold structured prediction, 32nd Conference on Neural Information Processing Systems (NIPS), Publisher: Massachusetts Institute of Technology Press, ISSN: 1049-5258
Structured prediction provides a general framework to deal with supervised problems where the outputs have semantically rich structure. While classical approaches consider finite, albeit potentially huge, output spaces, in this paper we discuss how structured prediction can be extended to a continuous scenario. Specifically, we study a structured prediction approach to manifold-valued regression. We characterize a class of problems for which the considered approach is statistically consistent and study how geometric optimization can be used to compute the corresponding estimator. Promising experimental results on both simulated and real data complete our study.
Denevi G, Ciliberto C, Stamos D, et al., 2018, Incremental Learning-to-Learn with Statistical Guarantees, 34th Conference on Uncertainty in Artificial Intelligence (UAI), Publisher: AUAI PRESS, Pages: 457-466
Denevi G, Ciliberto C, Stamos D, et al., 2018, Learning to learn around a common mean, Pages: 10169-10179, ISSN: 1049-5258
© 2018 Curran Associates Inc.All rights reserved. The problem of learning-to-learn (LTL) or meta-learning is gaining increasing attention due to recent empirical evidence of its effectiveness in applications. The goal addressed in LTL is to select an algorithm that works well on tasks sampled from a meta-distribution. In this work, we consider the family of algorithms given by a variant of Ridge Regression, in which the regularizer is the square distance to an unknown mean vector. We show that, in this setting, the LTL problem can be reformulated as a Least Squares (LS) problem and we exploit a novel meta-algorithm to efficiently solve it. At each iteration the meta-algorithm processes only one dataset. Specifically, it firstly estimates the stochastic LS objective function, by splitting this dataset into two subsets used to train and test the inner algorithm, respectively. Secondly, it performs a stochastic gradient step with the estimated value. Under specific assumptions, we present a bound for the generalization error of our meta-algorithm, which suggests the right splitting parameter to choose. When the hyper-parameters of the problem are fixed, this bound is consistent as the number of tasks grows, even if the sample size is kept constant. Preliminary experiments confirm our theoretical findings, highlighting the advantage of our approach, with respect to independent task learning.
Ciliberto C, Herbster M, Ialongo AD, et al., 2018, Quantum machine learning: a classical perspective, PROCEEDINGS OF THE ROYAL SOCIETY A-MATHEMATICAL PHYSICAL AND ENGINEERING SCIENCES, Vol: 474, ISSN: 1364-5021
Fanello SR, Valentin J, Kowdle A, et al., 2017, Low Compute and Fully Parallel Computer Vision with HashMatch, International Conference on Computer Vision, Pages: 3894-3903, ISSN: 1550-5499
© 2017 IEEE. Numerous computer vision problems such as stereo depth estimation, object-class segmentation and fore-ground/background segmentation can be formulated as per-pixel image labeling tasks. Given one or many images as input, the desired output of these methods is usually a spatially smooth assignment of labels. The large amount of such computer vision problems has lead to significant research efforts, with the state of art moving from CRF-based approaches to deep CNNs and more recently, hybrids of the two. Although these approaches have significantly advanced the state of the art, the vast majority has solely focused on improving quantitative results and are not designed for low-compute scenarios. In this paper, we present a new general framework for a variety of computer vision labeling tasks, called HashMatch. Our approach is designed to be both fully parallel, i.e. each pixel is independently processed, and low-compute, with a model complexity an order of magnitude less than existing CNN and CRF-based approaches. We evaluate HashMatch extensively on several problems such as disparity estimation, image retrieval, feature approximation and background subtraction, for which HashMatch achieves high computational efficiency while producing high quality results.
Ciliberto C, 2017, Connecting YARP to the Web with Yarp.js, FRONTIERS IN ROBOTICS AND AI, Vol: 4, ISSN: 2296-9144
Ciliberto C, Rudi A, Rosasco L, et al., 2017, Consistent Multitask Learning with Nonlinear Output Relations, 31st Conference on Neural Information Processing Systems (NIPS), Publisher: NEURAL INFORMATION PROCESSING SYSTEMS (NIPS), ISSN: 1049-5258
Camoriano R, Pasquale G, Ciliberto C, et al., 2017, Incremental robot learning of new objects with fixed update time, IEEE International Conference on Robotics and Automation, Pages: 3207-3214, ISSN: 1050-4729
© 2017 IEEE. We consider object recognition in the context of lifelong learning, where a robotic agent learns to discriminate between a growing number of object classes as it accumulates experience about the environment. We propose an incremental variant of the Regularized Least Squares for Classification (RLSC) algorithm, and exploit its structure to seamlessly add new classes to the learned model. The presented algorithm addresses the problem of having an unbalanced proportion of training examples per class, which occurs when new objects are presented to the system for the first time. We evaluate our algorithm on both a machine learning benchmark dataset and two challenging object recognition tasks in a robotic setting. Empirical evidence shows that our approach achieves comparable or higher classification performance than its batch counterpart when classes are unbalanced, while being significantly faster.
Higy B, Ciliberto C, Rosasco L, et al., 2016, Combining sensory modalities and exploratory procedures to improve haptic object recognition in robotics, IEEE-RAS International Conference on Humanoid Robots, Pages: 117-124, ISSN: 2164-0572
© 2016 IEEE. In this paper we tackle the problem of object recognition using haptic feedback from a robot holding and manipulating different objects. One of the main challenges in this setting is to understand the role of different sensory modalities (namely proprioception, object weight from F/T sensors and touch) and how to combine them to correctly discriminate different objects. We investigated these aspects by considering multiple sensory channels and different exploratory strategies to gather meaningful information regarding the object's physical properties. We propose a novel strategy to train a learning machine able to efficiently combine sensory modalities by first learning individual object features and then combine them in a single classifier. To evaluate our approach and compare it with previous methods we collected a dataset for haptic object recognition, comprising 11 objects that were held in the hands of the iCub robot while performing different exploration strategies. Results show that our strategy consistently outperforms previous approaches .
Jamali N, Ciliberto C, Rosasco L, et al., 2016, Active perception: Building objects' models using tactile exploration, IEEE-RAS International Conference on Humanoid Robots, Pages: 179-185, ISSN: 2164-0572
© 2016 IEEE. In this paper we present an efficient active learning strategy applied to the problem of tactile exploration of an object's surface. The method uses Gaussian process (GPs) classification to efficiently sample the surface of the object in order to reconstruct its shape. The proposed method iteratively samples the surface of the object, while, simultaneously constructing a probabilistic model of the object's surface. The probabilities in the model are used to guide the exploration. At each iteration, the estimate of the object's shape is used to slice the object in equally spaced intervals along the height of the object. The sampled locations are then labelled according to the interval in which their height falls. In its simple form, the data are labelled as belonging to the object and not belonging to the object: object and no-object, respectively. A GP classifier is trained to learn the object/no-object decision boundary. The next location to be sampled is selected at the classification boundary, in this way, the exploration is biased towards more informative areas. Complex features of the object's surface is captured by increasing the number of intervals as the number of sampled locations is increased. We validated our approach on six objects of different shapes using the iCub humanoid robot. Our experiments show that the method outperforms random selection and previous work based on GP regression by sampling more points on and near-the-boundary of the object.
Pasquale G, Ciliberto C, Rosasco L, et al., 2016, Object identification from few examples by improving the invariance of a deep convolutional neural network, IEEE International Conference on Intelligent Robots and Systems, Pages: 4904-4911, ISSN: 2153-0858
© 2016 IEEE. The development of reliable and robust visual recognition systems is a main challenge towards the deployment of autonomous robotic agents in unconstrained environments. Learning to recognize objects requires image representations that are discriminative to relevant information while being invariant to nuisances, such as scaling, rotations, light and background changes, and so forth. Deep Convolutional Neural Networks can learn such representations from large webcollected image datasets and a natural question is how these systems can be best adapted to the robotics context where little supervision is often available. In this work, we investigate different training strategies for deep architectures on a new dataset collected in a real-world robotic setting. In particular we show how deep networks can be tuned to improve invariance and discriminability properties and perform object identification tasks with minimal supervision.
Pasquale G, Mar T, Ciliberto C, et al., 2016, Enabling Depth-Driven Visual Attention on the iCub Humanoid Robot: Instructions for Use and New Perspectives, FRONTIERS IN ROBOTICS AND AI, Vol: 3, ISSN: 2296-9144
Ciliberto C, Rudi A, Rosasco L, 2016, A Consistent Regularization Approach for Structured Prediction, ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 29 (NIPS 2016), Vol: 29, ISSN: 1049-5258
Ciliberto C, Rosasco L, Villa S, 2015, Learning Multiple Visual Tasks while Discovering their Structure, 2015 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), Pages: 131-139, ISSN: 1063-6919
Ciliberto C, Mroueh Y, Poggio T, et al., 2015, Convex learning of multiple tasks and their structure, 32nd International Conference on Machine Learning, ICML 2015, Vol: 2, Pages: 1548-1557
Copyright © 2015 by the author(s). Reducing the amount of human supervision is a key problem in machine learning and a natural approach is that of exploiting the relations (structure) among different tasks. This is the idea at the core of multi-task learning. In this context a fundamental question is how to incorporate the tasks structure in the learning problem. We tackle this question by studying a general computational framework that allows to encode a-priori knowledge of the tasks structure in the form of a convex penalty; in this setting a variety of previously proposed methods can be recovered as special cases, including linear and non-linear approaches. Within this framework, we show that tasks and their structure can be efficiently learned considering a convex optimization problem that can be approached by means of block coordinate methods such as alternating minimization and for which we prove convergence to the global minimum.
Breschi GL, Ciliberto C, Nieus T, et al., 2015, Characterizing the Input-Output Function of the Olfactory-Limbic Pathway in the Guinea Pig, COMPUTATIONAL INTELLIGENCE AND NEUROSCIENCE, ISSN: 1687-5265
Ciliberto C, Fiorio L, Maggiali M, et al., 2014, Exploiting global force torque measurements for local compliance estimation in tactile arrays, IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Publisher: IEEE, Pages: 3994-3999, ISSN: 2153-0858
Fanello SR, Noceti N, Ciliberto C, et al., 2014, Ask the image: supervised pooling to preserve feature locality, 27th IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Publisher: IEEE, Pages: 851-858, ISSN: 1063-6919
Ciliberto C, Fanello SR, Santoro M, et al., 2013, On the impact of learning hierarchical representations for visual recognition in robotics, Pages: 3759-3764, ISSN: 2153-0858
Recent developments in learning sophisticated, hierarchical image representations have led to remarkable progress in the context of visual recognition. While these methods are becoming standard in modern computer vision systems, they are rarely adopted in robotics. The question arises of whether solutions, which have been primarily developed for image retrieval, can perform well in more dynamic and unstructured scenarios. In this paper we tackle this question performing an extensive evaluation of state of the art methods for visual recognition on a iCub robot. We consider the problem of classifying 15 different objects shown by a human demonstrator in a challenging Human-Robot Interaction scenario. The classification performance of hierarchical learning approaches are shown to outperform benchmark solutions based on local descriptors and template matching. Our results show that hierarchical learning systems are computationally efficient and can be used for real-time training and recognition of objects. © 2013 IEEE.
Fanello SR, Ciliberto C, Santoro M, et al., 2013, iCub World: Friendly Robots Help Building Good Vision Data-Sets, 2013 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS (CVPRW), Pages: 700-705, ISSN: 2160-7508
Fanello SR, Ciliberto C, Natale L, et al., 2013, Weakly Supervised Strategies for Natural Object Recognition in Robotics, IEEE International Conference on Robotics and Automation (ICRA), Publisher: IEEE, Pages: 4223-4229, ISSN: 1050-4729
Ciliberto C, Fanello SR, Natale L, et al., 2012, A Heteroscedastic Approach to Independent Motion Detection for Actuated Visual Sensors, 25th IEEE\RSJ International Conference on Intelligent Robots and Systems (IROS), Publisher: IEEE, Pages: 3907-3913, ISSN: 2153-0858
Ciliberto C, Smeraldi F, Natale L, et al., 2011, Online Multiple Instance Learning Applied to Hand Detection in a Humanoid Robot, IEEE/RSJ International Conference on Intelligent Robots and Systems, Publisher: IEEE, Pages: 1526-1532, ISSN: 2153-0858
Ciliberto C, Pattacini U, Natale L, et al., 2011, Reexamining Lucas-Kanade Method for Real-Time Independent Motion Detection: Application to the iCub Humanoid Robot, IEEE/RSJ International Conference on Intelligent Robots and Systems, Publisher: IEEE, Pages: 4154-4160, ISSN: 2153-0858
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.