Comparing support vector machines with Gaussian kernels to radialbasis function classifiers

The support vector (SV) machine is a novel type of learning machine, based on statistical learning theory, which contains polynomial classifiers, neural networks, and radial basis function (RBF) networks as special cases. In the RBF case, the SV algorithm automatically determines centers, weights, and threshold that minimize an upper bound on the expected test error. The present study is devoted to an experimental comparison of these machines with a classical approach, where the centers are determined by X-means clustering, and the weights are computed using error backpropagation. We consider three machines, namely, a classical RBF machine, an SV machine with Gaussian kernel, and a hybrid system with the centers determined by the SV method and the weights trained by error backpropagation. Our results show that on the United States postal service database of handwritten digits, the SV machine achieves the highest recognition accuracy, followed by the hybrid system. The SV approach is thus not only theoretically well-founded but also superior in a practical application     

Correction to the analysis of dielectric guides by transversemagnetic field finite element penalty function method with extensions

Recently a method to obtain the propagation constants of lossless dielectric waveguides using the Helmholtz equation with the finite element method and penalty function method was presented. The advantage of using this approach is that only one final eigenvalue matrix needs to be solved for only two components of the H-fields. We have determine that the results were obtained using an eigenvalue solver that did not account for the asymmetry in the final eigenvalue matrix. In this paper, we present the results of the same cases simulated using the correct eigenvalue solver, and the results obtained are in good agreement with previously published ones. We also show by simulation of appropriate cases, a high penalty factor is correlated to highly coupled modes, while weakly coupled modes may be correlated to small penalty factors. Finally, we have extended the penalty function method to include the complex case without the use of the perturbation method. The gain results obtained for a channel waveguide are in good agreement with previously published ones     

Fine tuning hole injection for high-performance polyfluorene-based blue emitting device by adjusting work function of anode via deposition of CFx layer with proper ionization potential on indium tin oxide

By introducing CF_x thin film as hole injection layer on top of indium tin oxide (ITO) anode via plasma polymerization of CHF₃, the device with poly(9,9-dioctylfluorene) (PFO) as emitting layer, ITO/CF_x(35 W)/PFO/CsF/Ca/Al, is prepared. At the optimal C/F atom ratio using the radio frequency power 35 W, the device performance is optimal having the maximum current efficiency 3.1 cd/A and maximum brightness 8400 cd/m². This is attributed to a better balance between hole and electron fluxes, resulting from a decrease in hole injection barrier as manifested by ultraviolet photoelectron spectroscopy and scanning surface potential microscopy.     

Detailed investigation of dependencies of photovoltaic performances of P3HT:PC61BM based solar cells on anodic work function modified by surface treatment of indium-tin-oxide electrode with benzenesulfonyl chloride derivatives

The photovoltaic (PV) characteristics of bulk-heterojunction (BHJ) solar cells based on poly(3-hexylthiophene) and [6,6]-phenyl-C₆₁-butyric acid methyl ester (P3HT:PC₆₁BM) were improved using indium-tin-oxide (ITO) anode electrodes modified chemically with CH₃O-, H-, Cl-, CF₃-, and NO₂-terminated benzenesulfonyl chlorides as a self-assembled monolayer (SAM). The ITO electrode surfaces were easily treated through the chemical modification of the reactive –SO₂Cl binding group, and the work function (WF) of the modified ITO was effectively changed depending on the permanent dipole moments introduced in the para-position of benzenesulfonyl chloride. We examined the correlation between the ITO WFs corrected by the change in the contact potential difference and the calculated dipole moments of the SAM models. Moreover, we examined the PV characteristics of the P3HT:PC₆₁BM based BHJ organic PV cells using the SAMs or poly(3,4-ethylenedioxythiophene):poly(styrene sulfonate) (PEDOT:PSS)-treated ITOs with different WFs lying within ±0.2 eV from the highest occupied molecular orbital (HOMO) level of P3HT. We found that the enhancement effect of the SAMs on the power-conversion efficiency (η_P) reached a maximum with Cl (η_P = 3.72%), and became larger than that of PEDOT:PSS (η_P = 3.62%). Two distinct J_sc dependencies, increasing and decreasing with the increasing WF of the anode ITO, were observed at higher and lower WFs than the HOMO level of the donor, respectively. Almost constant V_oc values (around 0.6 V) were observed with different SAM-modified ITOs, which suggested that Fermi level pinning was achieved by aligning the anode Fermi level and positive polaronic level of the donor polymer.