Determining a compact neural coding for a set of input stimuli is an issue that encompasses several biological memory mechanisms as well as various artificial neural network models. In particular, establishing the optimal network structure is still an open problem when dealing with unsupervised learning models. In this paper, we introduce a novel learning algorithm, named competitive repetition-suppression (CoRe) learning, inspired by a cortical memory mechanism called repetition suppression (RS). We show how such a mechanism is used, at various levels of the cerebral cortex, to generate compact neural representations of the visual stimuli. From the general CoRe learning model, we derive a clustering algorithm, named CoRe clustering, that can automatically estimate the unknown cluster number from the data without using a priori information concerning the input distribution. We illustrate how CoRe clustering, besides its biological plausibility, posses strong theoretical properties in terms of robustness to noise and outliers, and we provide an error function describing CoRe learning dynamics. Such a description is used to analyze CoRe relationships with the state-of-the art clustering models and to highlight CoRe similitude with rival penalized competitive learning (RPCL), showing how CoRe extends such a model by strengthening the rival penalization estimation by means of loss functions from robust statistics.

The paper introduces a robust clustering algorithm that can automatically determine the unknown cluster number from noisy data without any a-priori information. We show how our clustering algorithm can be derived from a general learning theory, named CoRe learning, that models a cortical memory mechanism called repetition suppression. Moreover, we describe CoRe clustering relationships with Rival Penalized Competitive Learning (RPCL), showing how CoRe extends this model by strengthening the rival penalization estimation by means of robust loss functions. Finally, we present the results of simulations concerning the unsupervised segmentation of noisy images.

This work presents an adaptive segmentation algorithm for endoscopic images. It is part of a complete system for robot-assisted endoscopy of the human sub-arachnoid spinal space. The role of the vision system is to provide a feedback for assisting the navigation of the endoscope and helping avoid damages to delicate tissues. Due to the presence of small blood vessels, nerves, and possible fibrosis, a multi-step approach has been followed for segmentation of the lumen (corresponding to free space for navigation) and the other tissues. Histogram analysis, together with blob analysis and a modified implementation of the convex hull algorithm bring to the isolation of the lumen; by means of adaptive thresholding nerves are isolated; thresholding on the hue and saturation helps in recognizing the vessels. A special condition of dirty lumen helps in managing doubtful situations. Experimental trials have been conducted on video streams from endoscopic explorations of animal (pig) spinal cord in-vivo. Experimental results show that membranes, vessels, nerves, and lumen are recognized in a reliable way, so as to contribute to robot-assisted endoscopy. The speed of the processing resulted compatible with an envisaged use in real tune support of endoscopic navigation.

We introduce a kernel for structured data, which is an extension of the Fisher kernel used for sequences. In our approach, we extract the Fisher score vectors from a Bayesian network, specifically a hidden tree Markov model, which can be constructed starting from the training data. Experiments on a QSPR (quantitative structure-property relationship) analysis, where instances are naturally represented as trees, allow a first test of the approach.

Studies of human motor control suggest that humans develop internal models of the arm during the execution of voluntary movements. In particular, the internal model consists of the inverse dynamic model of the musculoskeletal system and intervenes in the feedforward loop of the motor control system to improve reactivity and stability in rapid movements. In this paper, an interaction control scheme inspired by biological motor control is resumed, i.e. the coactivation-based compliance control in the joint space (Zollo, L, et al., 2003), and a feedforward module capable of online learning the manipulator inverse dynamics is presented. A novel recurrent learning paradigm is proposed which derives from an interesting functional equivalence between locally weighted regression networks and Takagi-Sugeno-Kang fuzzy systems. The proposed learning paradigm has been named recurrent locally weighted regression networks and strengthens the computational power of feedforward locally weighted regression networks. Simulation results are reported to validate the control scheme.

To overcome the problem of invariant pattern recognition, Simard, LeCun, and Denker (1993) proposed a successful nearest-neighbor approach based on tangent distance, attaining state-of-the-art accuracy. Since this approach needs great computational and memory effort, Hastie, Simard, and Säckinger (1995) proposed an algorithm (HSS) based on singular value decomposition (SVD), for the generation of nondiscriminant tangent models. In this article we propose a different approach, based on a gradient-descent constructive algorithm, called TD-Neuron, that develops discriminant models. We present as well comparative results of our constructive algorithm versus HSS and learning vector quantization (LVQ) algorithms. Specifically, we tested the HSS algorithm using both the original version based on the two-sided tangent distance and a new version based on the one-sided tangent distance. Empirical results over the NIST-3 database show that the TD-Neuron is superior to both SVD- and LVQ-based algorithms, since it reaches a better trade-off between error and rejection.

The capability of autonomously discovering relations between perceptual data and motor actions is crucial for the development of robust adaptive robotic systems intended to operate in a changing and unknown environment. In the case of robotic tactile perception, proper interaction between contact sensing and motor control is the basic step towards the execution of complex motor procedures such as grasping and manipulation. In this paper we propose an approach to the development of tactile-motor coordination in robotics, based on a neural model of the human tactile-motor system. The definition of such model is based on the features of biological systems as investigated by neuroscience. The autonomous development of tactile-motor coordination achieved through the implementation of the neural model is evaluated by experimental trials using a sensorised prosthetic hand and a robotic manipulator. The proposed neural network architecture linking changes in the sensed tactile pattern with the motor actions performed is described and experimental results are analysed and discussed.

This paper presents an integrated approach to tactile perception, both in terms of data acquisition and data interpretation. In humans, touch sensing is implemented through a number of different sensing elements embedded in the skin. The interpretation of perceived data to the level of detection of basic features, such as material, shape of surface, shape of contact, is achieved by integrating the different sensorial inputs at a low level, with no involvement of high level cognitive processes. The approach we propose in this paper follows this anthropomorphic model of tactile perception, by including, on one hand, a miniature fingertip integrating different sensors and, on the other hand, a parallel data interpretation module implemented through a fuzzy neural-network which processes all the different inputs at the same level. The paper describes the characteristics of the integrated fingertip sensor and of the neuro-fuzzy system, and discusses experimental results achieved during exploratory tasks on a set of common objects are discussed in detail.

Standard neural networks and statistical methods are usually believed to be inadequate when dealing with complex structures because of their feature-based approach. In fact, feature-based approaches usually fail to give satisfactory solutions because of the sensitivity of the approach to the a priori selection of the features, and the incapacity to represent any specific information on the relationships among the components of the structures. However, we show that neural networks can, in fact, represent and classify structured patterns. The key idea underpinning our approach is the use of the so called "generalized recursive neuron", which is essentially a generalization to structures of a recurrent neuron. By using generalized recursive neurons, all the supervised networks developed for the classification of sequences, such as backpropagation through time networks, real-time recurrent networks, simple recurrent networks, recurrent cascade correlation networks, and neural trees can, on the whole, be generalized to structures. The results obtained by some of the above networks (with generalized recursive neurons) on the classification of logic terms are presented.