Spatiotemporal pattern coding using Neural Fields: Optimal parameter estimation

نویسندگان

  • Mauricio Cerda
  • Bernard Girau
چکیده

In this work we present the parameter optimization of a distributed model for the classification of temporal sequences. The model we present codes information with a population of units, and it can be applied to discriminate between several spatiotemporal sequences. The implementation of this classification technique strongly depends on how the the connection weights are set according to the sequence we want to code. This “learning” phase depends on several parameters, for which we present a detailed analysis in this work. In the first part, we give some examples and we retrieve the model parameters in 1D by extending other existing studies. Later on we extend the analysis into 2D, proposing at the end of the report a strategy to code any given spatiotemporal sequence that can be described as a set local spatiotemporal trajectories. The derivation of parameters is performed analytically and numerical simulations are performed to verify our results. Key-words: neural fields, pattern classification, dynamical systems, computational neuroscience, spatiotemporal sequences. ∗ Université Henri Poincaré (UHP), LORIA, Cortex Project, Nancy, France; [email protected]. in ria -0 05 66 16 6, v er si on 1 15 F eb 2 01 1 Codage de motifs spatiotemporels par Champs neuronaux: Estimation optimale des paramètres Résumé : Dans ce rapport nous présentons l’optimisation du paramétrage d’un modéle distribué de discrimination de séquences temporelles. Le modèle que nous présentons code l’information au moyen d’une population d’unités, et il peut être appliqué à la discrimination de plusieurs séquences spatio-temporelles. L’implantation de cette technique de discrimination dépend fortement de la façon dont les poids de connexion sont fixés en fonction de la séquence que nous voulons coder. Cette phase d’apprentissage dépend de plusieurs paramètres, pour lesquels ce rapport présente une analyse détaillée. Dans la première partie, nous donnons quelques exemples et nous montrons comment déduire les paramètres du modèle dans sa forme 1D à partir d’autres travaux existants. Ensuite nous étendons l’analyse au cas 2D, en proposant à la fin du rapport une stratégie pour coder n’importe quelle séquence temporelle pouvant être décrite comme un ensemble de trajectoires spatio-temporelles locales. Les paramètres sont déterminés analytiquement et des simultations numériques sont réalisées pour vérifier nos résultats. Mots-clés : champs neuronaux, reconnaissance de motifs, systémes dynamique, neuroscience computationnelle, séquences spatio-temporelles in ria -0 05 66 16 6, v er si on 1 15 F eb 2 01 1 Optimal parameter estimation 3

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Spatiotemporal multi-resolution approximation of the Amari type neural field model

Neural fields are spatially continuous state variables described by integro-differential equations, which are well suited to describe the spatiotemporal evolution of cortical activations on multiple scales. Here we develop a multi-resolution approximation (MRA) framework for the integro-difference equation (IDE) neural field model based on semi-orthogonal cardinal B-spline wavelets. In this way...

متن کامل

Predictive Coding for Dynamic Visual Processing: Development of Functional Hierarchy in a Multiple Spatiotemporal Scales RNN Model

This letter proposes a novel predictive coding type neural network model, the predictive multiple spatiotemporal scales recurrent neural network (P-MSTRNN). The P-MSTRNN learns to predict visually perceived human whole-body cyclic movement patterns by exploiting multiscale spatiotemporal constraints imposed on network dynamics by using differently sized receptive fields as well as different tim...

متن کامل

Unsupervised and Supervised Clustering: The Mutual Information between Parameters and Observations

Recent works in parameter estimation and neural coding have demonstrated that optimal performance are related to the mutual information between parameters and data. We consider the mutual information in the case where the dependency in the parameter (a vector 8) of the conditional p.d.f. of each observation (a vector 0, is through the scalar product 8.~ only. We derive bounds and asymptotic beh...

متن کامل

Direct measurement of “suppression” in the LGN in the context of natural stimuli and its implications for visual coding

Some of the greatest advances in neural coding have occurred in the visual system, where the relationship between visual stimuli and the observed neuronal activity is relatively straightforward. However, experiments that probe this relationship in more “natural” contexts have revealed many higher order aspects of visual neuron responses, such as their precise timing down to milliseconds, adapta...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2011