Fast kernel classifiers with online and active learning

2005-09-01
Bordes, A
Ertekin Bolelli, Şeyda
Weston, J
Bottou, L
Very high dimensional learning systems become theoretically possible when training examples are abundant. The computing cost then becomes the limiting factor. Any efficient learning algorithm should at least take a brief look at each example. But should all examples be given equal attention?
JOURNAL OF MACHINE LEARNING RESEARCH

Suggestions

Fast and accurate solutions of scattering problems involving dielectric objects with moderate and low contrasts
Ergül, Özgür Salih (2007-08-31)
We consider the solution of electromagnetic scattering problems involving relatively large dielectric objects with moderate and low contrasts. Three-dimensional objects are discretized with Rao-Wilton-Glisson functions and the scattering problems are formulated with surface integral equations. The resulting dense matrix equations are solved iteratively by employing the multilevel fast multipole algorithm. We compare the accuracy and efficiency of the results obtained by employing various integral equations ...
Bridging the gap between variational homogenization results and two-scale asymptotic averaging techniques on periodic network structures
Kropat, Erik; Meyer-Nieberg, Silja; Weber, Gerhard Wilhelm (American Institute of Mathematical Sciences (AIMS), 2017)
In modern material sciences and multi-scale physics homogenization approaches provide a global characterization of physical systems that depend on the topology of the underlying microgeometry. Purely formal approaches such as averaging techniques can be applied for an identification of the averaged system. For models in variational form, two-scale convergence for network functions can be used to derive the homogenized model. The sequence of solutions of the variational microcsopic models and the correspondi...
The neural network technique - 1: a general exposition
Tulunay, Yurdanur; Tulunay, E; Senalp, ET (Elsevier BV, 2004-01-01)
Near earth space processes are highly complex and nonlinear and mathematical modeling based on first physical principals is usually difficult or impossible. For such cases data driven modeling methods are recommended to be used in parallel with mathematical modeling approach. Highly non-linear processes in the near-earth space are advantageously dealt with using data-driven modeling techniques in the neural network (NN) approach. The only basic requirement for its application is the availability of represen...
Deep learning algorithm applied to daily solar irradiation estimations
Akbaba, Erol C.; Yüce, Emre; Akinoglu, Bulent G. (2018-07-02)
Deep learning is applied in many research areas, and in many of them remarkable outcomes are attained compared to conventional methods. There are quite a number of studies also in the estimation of solar irradiation. Estimation of solar irradiation is vitally important for the design of solar energy systems. In this work, multi-layer perceptron (MLPs) method of deep learning is used to develop an estimation method for calculating horizontal daily solar irradiation and the results are compared with classical...
Efficient simulation-based discrete optimization
Guikema, SD; Davidson, RA; Ertuğrul, Zehra (2004-01-01)
In many practical applications of simulation it is desirable to optimize the levels of integer or binary variables that are inputs for the simulation model. In these cases, the objective function must often be estimated through an expensive simulation process, and the optimization problem is NP-hard, leading to a computationally difficult problem. We investigate efficient solution methods for this problem, and we propose an approach that reduces the number of runs of the simulation by using ridge regression...
Citation Formats
A. Bordes, Ş. Ertekin Bolelli, J. Weston, and L. Bottou, “Fast kernel classifiers with online and active learning,” JOURNAL OF MACHINE LEARNING RESEARCH, pp. 1579–1619, 2005, Accessed: 00, 2020. [Online]. Available: https://hdl.handle.net/11511/54343.