Results 1 -
9 of
9
Contingency perception and agency measure in visuo-motor spiking neural networks
- IEEE Trans. Auton. Mental Develop
, 2009
"... Abstract—Agency is the sense that I am the cause or author of a movement. Babies develop early this feeling by perceiving the contingency between afferent (sensor) and efferent (motor) infor-mation. A comparator model is hypothesized to be associated with many brain regions to monitor and simulate t ..."
Abstract
-
Cited by 5 (1 self)
- Add to MetaCart
(Show Context)
Abstract—Agency is the sense that I am the cause or author of a movement. Babies develop early this feeling by perceiving the contingency between afferent (sensor) and efferent (motor) infor-mation. A comparator model is hypothesized to be associated with many brain regions to monitor and simulate the concordance be-tween self-produced actions and their consequences. In this paper, we propose that the biological mechanism of spike timing-depen-dent plasticity, that synchronizes the neural dynamics almost ev-erywhere in the central nervous system, constitutes the perfect al-gorithm to detect contingency in sensorimotor networks. The co-herence or the dissonance in the sensorimotor information flow im-parts then the agency level. In a head-neck-eyes robot, we replicate three developmental experiments illustrating how particular per-ceptual experiences can modulate the overall level of agency inside the system; i.e., 1) by adding a delay between proprioceptive and visual feedback information, 2) by facing a mirror, and 3) a person. We show that the system learns to discriminate animated objects (self-image and other persons) from other type of stimuli. This sug-gests a basic stage representing the self in relation to others from low-level sensorimotor processes. We discuss then the relevance of our findings with neurobiological evidences and development psy-chological observations for developmental robots. Index Terms—Contingency detection, self-agency, sensorimotor integration, spiking neural networks.
Interactive object learning and recognition with multiclass support vector machines 169 0 Interactive object learning and recognition with multiclass support vector machines
"... ..."
(Show Context)
Acknowledgments
"... I want to thank my advisor, Prof. Giulio Sandini, for his invaluable support during these years. Besides the gratitude for his guidance, I would like to express my admiration for his eternal enthusiasm for research and his capacity to transmit it to those who, like me, have been privileged to work w ..."
Abstract
- Add to MetaCart
(Show Context)
I want to thank my advisor, Prof. Giulio Sandini, for his invaluable support during these years. Besides the gratitude for his guidance, I would like to express my admiration for his eternal enthusiasm for research and his capacity to transmit it to those who, like me, have been privileged to work with him. Special thanks go to Giorgio Metta for his help and patience, and particularly, for his friendship during these years. Thanks also to Lorenzo Natale who after a year has been able to satisfy his desire of revenge reviewing this thesis. I have to admit that he has done an excellent work, much better than the review I did for his thesis. I am eternally in debt with my wife Cristina for her infinite support and constant help. The most difficult part of this thesis has been to spend uncountable hours in front of the computer and not in her company. Finally, I would like to thank all my colleagues at the LIRA-Lab for their friendship and the professional and stimulating environment in which my thesis evolved.
Interactive
"... Interactive object learning and recognition with multiclass support vector machines 169 ..."
Abstract
- Add to MetaCart
(Show Context)
Interactive object learning and recognition with multiclass support vector machines 169
Towards a “chaotic ” smooth pursuit
"... Abstract—Real autonomous systems are very difficult to design, mainly due to the ever changing conditions of the environments where they are supposed to work. In the area of humanoid robotics these difficulties are increased not only because of the complexity of their mechanical structure, above all ..."
Abstract
- Add to MetaCart
(Show Context)
Abstract—Real autonomous systems are very difficult to design, mainly due to the ever changing conditions of the environments where they are supposed to work. In the area of humanoid robotics these difficulties are increased not only because of the complexity of their mechanical structure, above all because they are supposed to work under the same dynamic conditions as we humans do. Our approach for the creation of real autonomy in artificial systems is based on the use of nonlinear dynamical systems. The purpose of this research is to demonstrate the feasibility of using coupled chaotic systems within the area of cognitive developmental robotics. In our quest towards the design and implementation of a real self-adaptive autonomous cognitive architecture, we have decided to start with a simple application that will tell us how appropriate this approach can be for humanoid robots. Once an object appears in front of a camera, we demonstrate that the visual input is enough for the self-organization of the axes controlling the motion of a single eye, both in a virtual and a real platform. No learning or specific coding of the task is needed, which results in a very fast adaptation and robustness to perturbations. Another equally important goal of this research is the possibility of having new insights about how the coordination of multiple degrees of freedom emerges in human infants. I.
Eyes-Neck Coordination Using Chaos
"... Abstract. The increasing complexity of humanoid robots and their ex-pected performance in real dynamic environments demand an equally complex, autonomous and dynamic solution. Our approach for the cre-ation of real autonomy in artificial systems is based on the use of nonlin-ear dynamical systems. T ..."
Abstract
- Add to MetaCart
(Show Context)
Abstract. The increasing complexity of humanoid robots and their ex-pected performance in real dynamic environments demand an equally complex, autonomous and dynamic solution. Our approach for the cre-ation of real autonomy in artificial systems is based on the use of nonlin-ear dynamical systems. The purpose of this research is to demonstrate the feasibility of using coupled chaotic systems within the area of cogni-tive developmental robotics. Using a robotic head, we demonstrate that the visual input coming into the head’s eyes is enough for the self-organization of the axes controlling the motion of eyes and neck. No specific coding of the task is needed, which results in a very fast adaptation and robustness to perturbations. Another equally important goal of this research is the possibility of hav-ing new insights about how the coordination of multiple degrees of free-dom emerges in human infants. We show that the interaction between body and environment modifies the inner connections of the controlling network resulting in the emergence of a tracking behavior. 1
An iterative method for optimal resolution-constrained polar quantizer design
"... This paper addresses the problem of polar quantization optimization. Particularly, the aim of this investigation is to find the method for the optimal resolution-constrained polar quantizer design. The new iterative algorithm for determination of the optimal decision and represen-tation magnitude le ..."
Abstract
- Add to MetaCart
This paper addresses the problem of polar quantization optimization. Particularly, the aim of this investigation is to find the method for the optimal resolution-constrained polar quantizer design. The new iterative algorithm for determination of the optimal decision and represen-tation magnitude levels and algorithm for optimization of number of phase cells within each magnitude level, is proposed. At high rates, the new optimal polar quantizer outperforms the optimal polar com-pander for 0:2dB, while the more significant gain should be expected at lower rates. In this paper, in order to enable practical implementation of quantizer model, algorithm which transforms real values for the optimal numbers of phase cells within magnitude levels into integer ones is also proposed. Moreover, the approximate closed form of signal to quantization ratio (SQNR) is derived. Since circularly symmetric sources and complex presentation of signals arise in numer-ous applications, it can be concluded that the usage area of the suggested proposal is very wide (audio coding, image coding, spectral phase coding SPC, synthetic aperture radars systems SARs, coding of the discrete Fourier transform). It should be emphasized that in contrast to earlier work, where models have been de-signed under high-rate assumption, the obtained nonuniform unrestricted polar quantizer is optimal for all rates. 1