The Experts below are selected from a list of 83100 Experts worldwide ranked by ideXlab platform
Thomas Voegtlin - One of the best experts on this subject based on the ideXlab platform.
-
Recursive Principal Components Analysis
Neural Networks, 2005Co-Authors: Thomas VoegtlinAbstract:A recurrent linear network can be trained with Oja's constrained Hebbian learning rule. As a result, the network learns to represent the temporal context associated to its Input Sequence. The operation performed by the network is a generalization of Principal Components Analysis (PCA) to time-series, called Recursive PCA. The representations learned by the network are adapted to the temporal statistics of the Input. Moreover, Sequences stored in the network may be retrieved explicitly, in the reverse order of presentation, thus providing a straight-forward neural implementation of a logical stack.
Upinder S Bhalla - One of the best experts on this subject based on the ideXlab platform.
-
synaptic Input Sequence discrimination on behavioral timescales mediated by reaction diffusion chemistry in dendrites
eLife, 2017Co-Authors: Upinder S BhallaAbstract:Sequences of events are ubiquitous in sensory, motor, and cognitive function. Key computational operations, including pattern recognition, event prediction, and plasticity, involve neural discrimination of spatio-temporal Sequences. Here, we show that synaptically-driven reaction-diffusion pathways on dendrites can perform Sequence discrimination on behaviorally relevant time-scales. We used abstract signaling models to show that selectivity arises when Inputs at successive locations are aligned with, and amplified by, propagating chemical waves triggered by previous Inputs. We incorporated biological detail using sequential synaptic Input onto spines in morphologically, electrically, and chemically detailed pyramidal neuronal models based on rat data. Again, Sequences were recognized, and local channel modulation downstream of putative Sequence-triggered signaling could elicit changes in neuronal firing. We predict that dendritic Sequence-recognition zones occupy 5 to 30 microns and recognize time-intervals of 0.2 to 5 s. We suggest that this mechanism provides highly parallel and selective neural computation in a functionally important time range.
-
synaptic Input Sequence discrimination on behavioral time scales mediated by reaction diffusion chemistry in dendrites
bioRxiv, 2017Co-Authors: Upinder S BhallaAbstract:Sequences of events are ubiquitous in sensory, motor, and cognitive function. Key computational operations, including pattern recognition, event prediction, and plasticity, involve neural discrimination of spatio-temporal Sequences. Here we show that synaptically-driven reaction-diffusion pathways on dendrites can perform Sequence discrimination on behaviorally relevant time-scales. We used abstract signaling models to show that this selectivity arises when Inputs at successive locations are aligned with, and amplified by, propagating chemical waves triggered by previous Inputs. We incorporated biological detail using sequential synaptic Input onto spines in morphologically, electrically, and chemically detailed pyramidal neuronal models. Again, Sequences were recognized, and local channel modulation on the length-scale of Sequence Input could elicit changes in neuronal firing. We predict that dendritic Sequence-recognition zones occupy 5 to 20 microns and recognize time-intervals of 0.2 to 5s. We suggest that this mechanism provides highly parallel and selective neural computation in a functionally important time range.
M J Weinberger - One of the best experts on this subject based on the ideXlab platform.
-
universal discrete denoising known channel
International Symposium on Information Theory, 2003Co-Authors: Tsachy Weissman, Erik Ordentlich, G Seroussi, Sergio Verdu, M J WeinbergerAbstract:A discrete denoising algorithm estimates the Input Sequence to a discrete memoryless channel (DMC) based on the observation of the entire output Sequence. For the case in which the DMC is known and the quality of the reconstruction is evaluated with a given single-letter fidelity criterion, we propose a discrete denoising algorithm that does not assume knowledge of statistical properties of the Input Sequence. Yet, the algorithm is universal in the sense of asymptotically performing as well as the optimum denoiser that knows the Input Sequence distribution, which is only assumed to be stationary. Moreover, the algorithm is universal also in a semi-stochastic setting, in which the Input is an individual Sequence, and the randomness is due solely to the channel noise. The proposed denoising algorithm is practical, requiring a linear number of register-level operations and sublinear working storage size relative to the Input data length.
Robert M Hierons - One of the best experts on this subject based on the ideXlab platform.
-
an integrated search based approach for automatic testing from extended finite state machine efsm models
Information & Software Technology, 2011Co-Authors: Abdul Salam Kalaji, Robert M Hierons, Stephen SwiftAbstract:Context: The extended finite state machine (EFSM) is a modelling approach that has been used to represent a wide range of systems. When testing from an EFSM, it is normal to use a test criterion such as transition coverage. Such test criteria are often expressed in terms of transition paths (TPs) through an EFSM. Despite the popularity of EFSMs, testing from an EFSM is difficult for two main reasons: path feasibility and path Input Sequence generation. The path feasibility problem concerns generating paths that are feasible whereas the path Input Sequence generation problem is to find an Input Sequence that can traverse a feasible path. Objective: While search-based approaches have been used in test automation, there has been relatively little work that uses them when testing from an EFSM. In this paper, we propose an integrated search-based approach to automate testing from an EFSM. Method: The approach has two phases, the aim of the first phase being to produce a feasible TP (FTP) while the second phase searches for an Input Sequence to trigger this TP. The first phase uses a Genetic Algorithm whose fitness function is a TP feasibility metric based on dataflow dependence. The second phase uses a Genetic Algorithm whose fitness function is based on a combination of a branch distance function and approach level. Results: Experimental results using five EFSMs found the first phase to be effective in generating FTPs with a success rate of approximately 96.6%. Furthermore, the proposed Input Sequence generator could trigger all the generated feasible TPs (success rate=100%). Conclusion: The results derived from the experiment demonstrate that the proposed approach is effective in automating testing from an EFSM.
-
Input Sequence generation for testing of communicating finite state machines cfsms
Genetic and Evolutionary Computation Conference, 2004Co-Authors: Karnig Derderian, Robert M Hierons, Mark Harman, Qiang GuoAbstract:Finite State Machines (FSMs) have been used to model systems in different areas like sequential circuits, software development and communication protocols [1]. FSMs have been an effective method of modelling because a variety of techniques and automated tools exist that work with them.
M Verhaegen - One of the best experts on this subject based on the ideXlab platform.
-
subspace identification of bilinear systems using a dedicated Input Sequence
Conference on Decision and Control, 2007Co-Authors: J W Van Wingerden, M VerhaegenAbstract:A novel identification scheme for bilinear systems is proposed. Piecewise constant Input Sequences are used to overcome the curse of dimensionality of traditional bilinear subspace identification methods. This dedicated Input Sequence allows for the identification of stationary Input models using traditional subspace identification techniques. Inherent to sub-space identification these models are identified in a different state basis. A common state basis can be found by setting up an intersection problem combined with a number of linear relations. The solution of this linear problem can be used to estimate the system matrices. The working of the algorithm is illustrated with a simulation example.
-
CDC - Subspace identification of bilinear systems using a dedicated Input Sequence
2007 46th IEEE Conference on Decision and Control, 2007Co-Authors: J W Van Wingerden, M VerhaegenAbstract:A novel identification scheme for bilinear systems is proposed. Piecewise constant Input Sequences are used to overcome the curse of dimensionality of traditional bilinear subspace identification methods. This dedicated Input Sequence allows for the identification of stationary Input models using traditional subspace identification techniques. Inherent to sub-space identification these models are identified in a different state basis. A common state basis can be found by setting up an intersection problem combined with a number of linear relations. The solution of this linear problem can be used to estimate the system matrices. The working of the algorithm is illustrated with a simulation example.
