F. Orabona - Academia.edu (original) (raw)
Papers by F. Orabona
I show a simple expression of the Mill's ratio of the Student's t-Distribution. I use it to prove... more I show a simple expression of the Mill's ratio of the Student's t-Distribution. I use it to prove Conjecture 1 in Peter Auer, Nicolò Cesa-Bianchi, and Paul Fischer. Finite-time analysis of the multiarmed bandit problem. Mach. Learn., 47(2-3): 235-256, May 2002.
Stochastic gradient descent algorithms for training linear and kernel predictors are gaining more... more Stochastic gradient descent algorithms for training linear and kernel predictors are gaining more and more importance, thanks to their scalability. While various methods have been proposed to speed up their convergence, the model selection phase is often ignored. In fact, in theoretical works most of the time assumptions are made, for example, on the prior knowledge of the norm of the optimal solution, while in the practical world validation methods remain the only viable approach. In this paper, we propose a new kernel-based stochastic gradient descent algorithm that performs model selection while training, with no parameters to tune, nor any form of cross-validation. The algorithm builds on recent advancement in online learning theory for unconstrained settings, to estimate over time the right regularization in a datadependent way. Optimal rates of convergence are proved under standard smoothness assumptions on the target function, using the range space of the fractional integral operator associated with the kernel.
We study algorithms for online linear optimization in Hilbert spaces, focusing on the case where ... more We study algorithms for online linear optimization in Hilbert spaces, focusing on the case where the player is unconstrained. We develop a novel characterization of a large class of minimax algorithms, recovering, and even improving, several previous results as immediate corollaries. Moreover, using our tools, we develop an algorithm that provides a regret bound of O U T log(U √ T log 2 T + 1) where U is the L2 norm of a comparator, and both T and U are unknown to the player. This bound is optimal up to √ log log T terms. When T is known, we derive an algorithm, whose regret bound is optimal up to constant terms. For both the known and unknown T case, a Normal approximation to the conditional value of the game proves to be the key analysis tool. * Both authors thank Jacob Abernethy for insightful feedback on this work.
We propose a general framework to online learning for classification problems with time-varying p... more We propose a general framework to online learning for classification problems with time-varying potential functions in the adversarial setting. This framework allows to design and prove relative mistake bounds for any generic loss function. The mistake bounds can be specialized for the hinge loss, allowing to recover and improve the bounds of known online classification algorithms. By optimizing the general bound we derive a new online classification algorithm, called NAROW, that hybridly uses adaptive-and fixed-second order information. We analyze the properties of the algorithm and illustrate its performance using synthetic dataset.
Lecture Notes in Computer Science, 2015
ABSTRACT We design algorithms for online linear optimization that have optimal regret and at the ... more ABSTRACT We design algorithms for online linear optimization that have optimal regret and at the same time do not need to know any upper or lower bounds on the norm of the loss vectors. We achieve adaptiveness to norms of loss vectors by scale invariance, i.e., our algorithms make exactly the same decisions if the sequence of loss vectors is multiplied by any positive constant. Our algorithms work for any decision set, bounded or unbounded. For unbounded decisions sets, these are the first truly adaptive algorithms for online linear optimization.
Lecture Notes in Computer Science, 2015
Lecture Notes in Computer Science, 2008
One of the first steps of any visual system is that of locating suitable interest points, 'salien... more One of the first steps of any visual system is that of locating suitable interest points, 'salient regions', in the scene, to detect events, and eventually to direct gaze toward these locations. In the last few years, object-based visual attention models have received an increasing interest in computational neuroscience and in computer vision, the problem, in this case, being that of creating a model of 'objecthood' that eventually guides a saliency mechanism. We present here an model of visual attention based on the definition of 'proto-objects' and show its instantiation on a humanoid robot. Moreover we propose a biological plausible way to learn certain Gestalt rules that can lead to proto-objects.
2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Workshops, 2005
One of the first steps of any visual system is that of locating suitable interest points, "salien... more One of the first steps of any visual system is that of locating suitable interest points, "salient regions", in the scene, to detect events, and eventually to direct gaze toward these locations. In the last few years, object-based visual attention models have received an increasing interest in the literature, the problem, in this case, being that of creating a model of "objecthood" that eventually guides a saliency mechanism. We propose here an object-based model of visual attention and show its instantiation on a humanoid robot. The robot employs action to learn and define its own concept of objecthood.
2009 IEEE International Conference on Robotics and Automation, 2009
The state-of-the-art in control of hand prosthetics is far from optimal. The main control interfa... more The state-of-the-art in control of hand prosthetics is far from optimal. The main control interface is represented by surface electromyography (EMG): the activation potentials of the remnants of large muscles of the stump are used in a nonnatural way to control one or, at best, two degrees-of-freedom. This has two drawbacks: first, the dexterity of the prosthesis is limited, leading to poor interaction with the environment; second, the patient undergoes a long training time. As more dexterous hand prostheses are put on the market, the need for a finer and more natural control arises. Machine learning can be employed to this end. A desired feature is that of providing a pre-trained model to the patient, so that a quicker and better interaction can be obtained.
Conference proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2006
Retinal prosthesis represents the best near-term hope for individuals with chronic blinding disea... more Retinal prosthesis represents the best near-term hope for individuals with chronic blinding disease of the outer retina. However the small number of stimulating electrodes produces a poor, low resolution image. We propose a new preprocessing method for epi-retinal implants and validate it through a novel simulation of the implanted blind perception. Twenty-one normally sighted, untrained subjects performed a face recognition test. Three different electrodes grids were simulated: rectangular, hexagonal and log-polar. The results show that the proposed pre-processing method has a good and statistically significant performance improvement.
Progress in brain research, 2007
This paper describes a developmental approach to the design of a humanoid robot. The robot, equip... more This paper describes a developmental approach to the design of a humanoid robot. The robot, equipped with initial perceptual and motor competencies, explores the "shape" of its own body before devoting its attention to the external environment. The initial form of sensorimotor coordination consists of a set of explorative motor behaviors coupled to visual routines providing a bottom-up sensory-driven attention system. Subsequently, development leads the robot from the construction of a "body schema" to the exploration of the world of objects. The "body schema" allows controlling the arm and hand to reach and touch objects within the robot's workspace. Eventually, the interaction between the environment and the robot's body is exploited to acquire a visual model of the objects the robot encounters which can then be used to guide a top-down attention system.
The maximum a-posteriori (MAP) perturbation framework has emerged as a useful approach for infere... more The maximum a-posteriori (MAP) perturbation framework has emerged as a useful approach for inference and learning in high dimensional complex models. By maximizing a randomly perturbed potential function, MAP perturbations generate unbiased samples from the Gibbs distribution. Unfortunately, the computational cost of generating so many high-dimensional random variables can be prohibitive. More efficient algorithms use sequential sampling strategies based on the expected value of low dimensional MAP perturbations. This paper develops new measure concentration inequalities that bound the number of samples needed to estimate such expected values. Applying the general result to MAP perturbations can yield a more efficient algorithm to approximate sampling from the Gibbs distribution. The measure concentration result is of general interest and may be applicable to other areas involving expected estimations.
I show a simple expression of the Mill's ratio of the Student's t-Distribution. I use it to prove... more I show a simple expression of the Mill's ratio of the Student's t-Distribution. I use it to prove Conjecture 1 in Peter Auer, Nicolò Cesa-Bianchi, and Paul Fischer. Finite-time analysis of the multiarmed bandit problem. Mach. Learn., 47(2-3): 235-256, May 2002.
Stochastic gradient descent algorithms for training linear and kernel predictors are gaining more... more Stochastic gradient descent algorithms for training linear and kernel predictors are gaining more and more importance, thanks to their scalability. While various methods have been proposed to speed up their convergence, the model selection phase is often ignored. In fact, in theoretical works most of the time assumptions are made, for example, on the prior knowledge of the norm of the optimal solution, while in the practical world validation methods remain the only viable approach. In this paper, we propose a new kernel-based stochastic gradient descent algorithm that performs model selection while training, with no parameters to tune, nor any form of cross-validation. The algorithm builds on recent advancement in online learning theory for unconstrained settings, to estimate over time the right regularization in a datadependent way. Optimal rates of convergence are proved under standard smoothness assumptions on the target function, using the range space of the fractional integral operator associated with the kernel.
We study algorithms for online linear optimization in Hilbert spaces, focusing on the case where ... more We study algorithms for online linear optimization in Hilbert spaces, focusing on the case where the player is unconstrained. We develop a novel characterization of a large class of minimax algorithms, recovering, and even improving, several previous results as immediate corollaries. Moreover, using our tools, we develop an algorithm that provides a regret bound of O U T log(U √ T log 2 T + 1) where U is the L2 norm of a comparator, and both T and U are unknown to the player. This bound is optimal up to √ log log T terms. When T is known, we derive an algorithm, whose regret bound is optimal up to constant terms. For both the known and unknown T case, a Normal approximation to the conditional value of the game proves to be the key analysis tool. * Both authors thank Jacob Abernethy for insightful feedback on this work.
We propose a general framework to online learning for classification problems with time-varying p... more We propose a general framework to online learning for classification problems with time-varying potential functions in the adversarial setting. This framework allows to design and prove relative mistake bounds for any generic loss function. The mistake bounds can be specialized for the hinge loss, allowing to recover and improve the bounds of known online classification algorithms. By optimizing the general bound we derive a new online classification algorithm, called NAROW, that hybridly uses adaptive-and fixed-second order information. We analyze the properties of the algorithm and illustrate its performance using synthetic dataset.
Lecture Notes in Computer Science, 2015
ABSTRACT We design algorithms for online linear optimization that have optimal regret and at the ... more ABSTRACT We design algorithms for online linear optimization that have optimal regret and at the same time do not need to know any upper or lower bounds on the norm of the loss vectors. We achieve adaptiveness to norms of loss vectors by scale invariance, i.e., our algorithms make exactly the same decisions if the sequence of loss vectors is multiplied by any positive constant. Our algorithms work for any decision set, bounded or unbounded. For unbounded decisions sets, these are the first truly adaptive algorithms for online linear optimization.
Lecture Notes in Computer Science, 2015
Lecture Notes in Computer Science, 2008
One of the first steps of any visual system is that of locating suitable interest points, 'salien... more One of the first steps of any visual system is that of locating suitable interest points, 'salient regions', in the scene, to detect events, and eventually to direct gaze toward these locations. In the last few years, object-based visual attention models have received an increasing interest in computational neuroscience and in computer vision, the problem, in this case, being that of creating a model of 'objecthood' that eventually guides a saliency mechanism. We present here an model of visual attention based on the definition of 'proto-objects' and show its instantiation on a humanoid robot. Moreover we propose a biological plausible way to learn certain Gestalt rules that can lead to proto-objects.
2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Workshops, 2005
One of the first steps of any visual system is that of locating suitable interest points, "salien... more One of the first steps of any visual system is that of locating suitable interest points, "salient regions", in the scene, to detect events, and eventually to direct gaze toward these locations. In the last few years, object-based visual attention models have received an increasing interest in the literature, the problem, in this case, being that of creating a model of "objecthood" that eventually guides a saliency mechanism. We propose here an object-based model of visual attention and show its instantiation on a humanoid robot. The robot employs action to learn and define its own concept of objecthood.
2009 IEEE International Conference on Robotics and Automation, 2009
The state-of-the-art in control of hand prosthetics is far from optimal. The main control interfa... more The state-of-the-art in control of hand prosthetics is far from optimal. The main control interface is represented by surface electromyography (EMG): the activation potentials of the remnants of large muscles of the stump are used in a nonnatural way to control one or, at best, two degrees-of-freedom. This has two drawbacks: first, the dexterity of the prosthesis is limited, leading to poor interaction with the environment; second, the patient undergoes a long training time. As more dexterous hand prostheses are put on the market, the need for a finer and more natural control arises. Machine learning can be employed to this end. A desired feature is that of providing a pre-trained model to the patient, so that a quicker and better interaction can be obtained.
Conference proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2006
Retinal prosthesis represents the best near-term hope for individuals with chronic blinding disea... more Retinal prosthesis represents the best near-term hope for individuals with chronic blinding disease of the outer retina. However the small number of stimulating electrodes produces a poor, low resolution image. We propose a new preprocessing method for epi-retinal implants and validate it through a novel simulation of the implanted blind perception. Twenty-one normally sighted, untrained subjects performed a face recognition test. Three different electrodes grids were simulated: rectangular, hexagonal and log-polar. The results show that the proposed pre-processing method has a good and statistically significant performance improvement.
Progress in brain research, 2007
This paper describes a developmental approach to the design of a humanoid robot. The robot, equip... more This paper describes a developmental approach to the design of a humanoid robot. The robot, equipped with initial perceptual and motor competencies, explores the "shape" of its own body before devoting its attention to the external environment. The initial form of sensorimotor coordination consists of a set of explorative motor behaviors coupled to visual routines providing a bottom-up sensory-driven attention system. Subsequently, development leads the robot from the construction of a "body schema" to the exploration of the world of objects. The "body schema" allows controlling the arm and hand to reach and touch objects within the robot's workspace. Eventually, the interaction between the environment and the robot's body is exploited to acquire a visual model of the objects the robot encounters which can then be used to guide a top-down attention system.
The maximum a-posteriori (MAP) perturbation framework has emerged as a useful approach for infere... more The maximum a-posteriori (MAP) perturbation framework has emerged as a useful approach for inference and learning in high dimensional complex models. By maximizing a randomly perturbed potential function, MAP perturbations generate unbiased samples from the Gibbs distribution. Unfortunately, the computational cost of generating so many high-dimensional random variables can be prohibitive. More efficient algorithms use sequential sampling strategies based on the expected value of low dimensional MAP perturbations. This paper develops new measure concentration inequalities that bound the number of samples needed to estimate such expected values. Applying the general result to MAP perturbations can yield a more efficient algorithm to approximate sampling from the Gibbs distribution. The measure concentration result is of general interest and may be applicable to other areas involving expected estimations.