In a previous paper, we have proposed a visuo-motor control architecture, which we called MEP, or... more In a previous paper, we have proposed a visuo-motor control architecture, which we called MEP, oriented to the execution of Goal Ori-ented actions (GO-action). Here the expression "Goal Oriented action" is used to denote a series of prehension movements that relate body parts of the subject to a three-dimensional object. MEP architecture is based on a biologically inspired expected perception mechanism. In this paper we discuss some issues concerning the implementation of the proposed architecture. In particular, we focus our attention on the implementation of the expected perception mechanism. To this end, we argue some basic hypothesis regarding the semantic segmentation of GO-action and their observer independence representation. We give a preliminary account of how the plausibility of such hypothesis can be fulfilled and tested in experimental settings.
In this paper we propose a general framework to characterize and solve the optimization problems ... more In this paper we propose a general framework to characterize and solve the optimization problems underlying a large class of sparsity based regularization algorithms. More precisely, we study the minimization of learning functionals that are sums of a differentiable data term and a convex non differentiable penalty. These latter penalties have recently become popular in machine learning since they allow to enforce various kinds of sparsity properties in the solution. Leveraging on the theory of Fenchel duality and subdifferential calculus, we derive explicit optimality conditions for the regularized solution and propose a general iterative projection algorithm whose convergence to the optimal solution can be proved. The generality of the framework is illustrated, considering several examples of regularization schemes, including l1 regularization (and several variants), multiple kernel learning and multi-task learning. Finally, some features of the proposed framework are empirically ...
... used for behavioral analysis (through the so called ARMA models [5]), diffusion processes, gr... more ... used for behavioral analysis (through the so called ARMA models [5]), diffusion processes, graphs ...Motion and appearance information are measured, stored, and updated indepen-dently, but jointly ... cases it may be trou-blesome to consider them as probabilistic generators of ...
Nijhawan argues convincingly that predictive mechanisms are pervasive in the central nervous syst... more Nijhawan argues convincingly that predictive mechanisms are pervasive in the central nervous system (CNS). However, scientific understanding of visual prediction requires one to formulate empirically testable neurophysiological models. The author's suggestions in this direction are to be evaluated on the basis of more realistic experimental methodologies and more plausible assumptions on the hierarchical character of the human visual cortex.
ABSTRACT The understanding of behaviours is important for many monitoring tasks. In this paper we... more ABSTRACT The understanding of behaviours is important for many monitoring tasks. In this paper we propose an unsupervised method to gather knowledge of common behaviours in a scene from long-time observation. We adopt a spectral clustering approach and combine it with a string-based event description that allows us to extract repetitive behaviours from sets of events. An experimental analysis, carried out on both synthetic and real video-surveillance data, confirms the appropriateness of the approach as a versatile knowledge discovery method from cluttered temporal data.
In a previous paper, we have proposed a visuo-motor control architecture, which we called MEP, or... more In a previous paper, we have proposed a visuo-motor control architecture, which we called MEP, oriented to the execution of Goal Ori-ented actions (GO-action). Here the expression "Goal Oriented action" is used to denote a series of prehension movements that relate body parts of the subject to a three-dimensional object. MEP architecture is based on a biologically inspired expected perception mechanism. In this paper we discuss some issues concerning the implementation of the proposed architecture. In particular, we focus our attention on the implementation of the expected perception mechanism. To this end, we argue some basic hypothesis regarding the semantic segmentation of GO-action and their observer independence representation. We give a preliminary account of how the plausibility of such hypothesis can be fulfilled and tested in experimental settings.
In this paper we propose a general framework to characterize and solve the optimization problems ... more In this paper we propose a general framework to characterize and solve the optimization problems underlying a large class of sparsity based regularization algorithms. More precisely, we study the minimization of learning functionals that are sums of a differentiable data term and a convex non differentiable penalty. These latter penalties have recently become popular in machine learning since they allow to enforce various kinds of sparsity properties in the solution. Leveraging on the theory of Fenchel duality and subdifferential calculus, we derive explicit optimality conditions for the regularized solution and propose a general iterative projection algorithm whose convergence to the optimal solution can be proved. The generality of the framework is illustrated, considering several examples of regularization schemes, including l1 regularization (and several variants), multiple kernel learning and multi-task learning. Finally, some features of the proposed framework are empirically ...
... used for behavioral analysis (through the so called ARMA models [5]), diffusion processes, gr... more ... used for behavioral analysis (through the so called ARMA models [5]), diffusion processes, graphs ...Motion and appearance information are measured, stored, and updated indepen-dently, but jointly ... cases it may be trou-blesome to consider them as probabilistic generators of ...
Nijhawan argues convincingly that predictive mechanisms are pervasive in the central nervous syst... more Nijhawan argues convincingly that predictive mechanisms are pervasive in the central nervous system (CNS). However, scientific understanding of visual prediction requires one to formulate empirically testable neurophysiological models. The author's suggestions in this direction are to be evaluated on the basis of more realistic experimental methodologies and more plausible assumptions on the hierarchical character of the human visual cortex.
ABSTRACT The understanding of behaviours is important for many monitoring tasks. In this paper we... more ABSTRACT The understanding of behaviours is important for many monitoring tasks. In this paper we propose an unsupervised method to gather knowledge of common behaviours in a scene from long-time observation. We adopt a spectral clustering approach and combine it with a string-based event description that allows us to extract repetitive behaviours from sets of events. An experimental analysis, carried out on both synthetic and real video-surveillance data, confirms the appropriateness of the approach as a versatile knowledge discovery method from cluttered temporal data.
Uploads
Papers by Matteo Santoro