A Learning Algorithm Based on PSO and L-M for Parity Problem

Yang, Guangyou; Zhang, Daode; Hu, Xinyu

doi:10.5772/46973

Cited by 11 publications

(13 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…An earlier work employing the original PSO reported the resolution of XOR problem with just 9 epochs on average [25]. In another work, alternative PSO versions were able to solve the XOR problem in dozens of evaluations, while the basic PSO was not even able to converge [26]. In [27], a variety of backpropagation methods were related to results ranging from 24 to 3495 epochs, and the author highlighted that the stopping criterion used may greatly influence the obtained number of epochs.…”

Section: Resultsmentioning

confidence: 99%

Error measures based on winner-takes-all and the performance of multilayer perceptron classifier

Santos¹,

Santos²

2016

Anais Do 11. Congresso Brasileiro De Inteligência Computacional

View full text Add to dashboard Cite

Abstract-The squared error is a measure commonly employed for training neural networks. Alternative objective functions, directed for classification tasks, may be relevant for training neural networks classifiers. The present paper discusses why some types of objective function, used in training neural networks for classification, have shown better performance compared to the usual mean squared error (MSE). The study deals with the concept of winner-takes-all (WTA). For the condition of a trained network, it is demonstrated through a variance inequality that the application of the smooth WTA criterion (softmax) in the output estimates made by a multilayer perceptron (MLP) classifier tends to reduce the variance of such estimates. However, as the softmax approximates the abrupt WTA criterion, there is no variance reduction due to such inequality. On the other hand, considering a network to be trained, error measures based on the composition of MSE with WTA are defined to serve as objective function of particle swarm optimizer in training a MLP for solving logic gates classification. Experimental results point out that the steeper is the WTA criterion, the faster is reaching a solution for the AND gate, moreover, the functions depending on a steeper WTA were superior to the simple MSE. Curiously, in the case of XOR problem, the solution was more often reached when using the pure MSE, whereas most objective functions based on the WTA criterion were not able to achieve a solution before the maximum number of evaluations.

show abstract

Section: Resultsmentioning

confidence: 99%

Error measures based on winner-takes-all and the performance of multilayer perceptron classifier

Santos¹,

Santos²

2016

Anais Do 11. Congresso Brasileiro De Inteligência Computacional

View full text Add to dashboard Cite

show abstract

“…can also be open for low values of mÂ and mĥ 1 , but their decay branching ratios are relatively small and can be neglected [8].…”

Section: The Dark Matter In the Lrth Modelmentioning

confidence: 99%

“…The collider phenomenology of the LRTH model has been studied intensively for example in Refs. [3][4][5][6][7][8].…”

Section: Introductionmentioning

confidence: 99%

Constraining dark matter in the LRTH model with latest LHC, XENON100 and LUX data

Liu¹,

Xiao²

2015

J. Phys. G: Nucl. Part. Phys.

View full text Add to dashboard Cite

In the left-right twin Higgs (LRTH) model, the neutralŜ is a candidate for weakly interacting massive particle (WIMP) dark matter (DM). If its mass is lighter than half of the SM-like Higgs boson h, the new invisible decay h →ŜŜ will become open. In this paper, we examine the status of a light dark matter (Ŝ ) under current experimental constraints including the latest LHC Higgs data, the XENON100 and LUX limit on the dark matter scattering off the nucleon. The following observations have been obtained: (i) The current ATLAS (CMS) measurements of R γγ can exclude the invisible Higgs branching ratio Br inv larger than 34% (48%) at 2σ level; (ii) the Global fits to the latest LHC and Tevatron Higgs data provide a stronger constraint: Br inv < 20% (30%) at 2σ (3σ) level, which could be tested in the LHC experiments; (iii) for the spin-independent scattering cross section off the nucleon, the recent XENON100 (LUX) data can exclude the invisible decay rate larger than 50% (25%); and (iv) the results of direct DM searches with LUX can give strong constraint on the viable parameter space of {g hŜŜ , mŜ} in this LRTH model.

show abstract

“…For example Shafiee et al (2013) have investigated the effect of three different dictionary learning methods on SRC method. Performance of SRC method affects by number of training data, they used Metaface dictionary learning (Yang et al, 2010b), Fisher Discriminative Dictionary Learning (FDDL) (Yang et al, 2011a), Sparse Modelling Representative Selection (SMRS) (Elhamifar et al, 2012) to obtain compact representation of training data. They were able to show that FDDL method performs a high recognition accuracy compare to other methods although SMRS method requires a much smaller learning time compare to others.…”

Section: Dictionary Learningmentioning

confidence: 99%

Learning of Graph Compressed Dictionaries for Sparse Representation Classification

Nourbakhsh

Granger

2016

Proceedings of the 5th International Conference on Pattern Recognition Applications and Methods

View full text Add to dashboard Cite

Abstract:Despite the limited target data available to design face models in video surveillance applications, many faces of non-target individuals may be captured in operational environments, and over multiple cameras, to improve robustness to variations. This paper focuses on Sparse Representation Classification (SRC) techniques that are suitable for the design of still-to-video FR systems based on under-sampled dictionaries. The limited reference data available during enrolment is complemented by an over-complete external dictionary that is formed with an abundance of faces from non-target individuals. In this paper, the Graph-Compressed Dictionary Learning (GCDL) technique is proposed to learn compact auxiliary dictionaries for SRC. GCDL is based on matrix factorization, and allows to maintain a high level of accuracy with compressed dictionaries because it exploits structural information to represent intra-class variations. Graph factorization compression has been shown to efficiently compress data, and can therefore rapidly construct compressed dictionaries. Accuracy and efficiency of the proposed technique is assessed and compared to reference sparse coding and dictionary learning technique using videos from the CAS-PEAL database. GCDL is shown to provide fast matching and adaptation of compressed dictionaries to new reference faces from the video surveillance environments.

show abstract

A Learning Algorithm Based on PSO and L-M for Parity Problem

Cited by 11 publications

References 1 publication

Error measures based on winner-takes-all and the performance of multilayer perceptron classifier

Error measures based on winner-takes-all and the performance of multilayer perceptron classifier

Constraining dark matter in the LRTH model with latest LHC, XENON100 and LUX data

Learning of Graph Compressed Dictionaries for Sparse Representation Classification

Contact Info

Product

Resources

About