Maxillary protraction using the 2-hinged expander, a repetitive weekly protocol of Alternate Rapid Maxillary Expansions and Constrictions, and intraoral protraction springs is most effective, with stable results at 2-year follow-up.

No abstract

High-and low-frequency pulse current ͑PC͒ and pulse-reverse current ͑PR͒ were employed to the plated-through-hole ͑PTH͒ process. To understand the role of additives in PC and PR, chloride ions, polyethylene glycol ͑PEG͒, and 3-mercapto-1propanesulfonate ͑MPS͒ were used as additives and compared to baths without additives. Linear sweep voltammetry, cyclic voltammetry, and impedance analysis were employed to characterize this system on a rotating disk electrode. In the bath without additives, it was found that deposition uniformity in PTH with high-frequency PC and PR at 20 mA/cm 2 average current density were better than that with low-frequency PC and PR, even better than that of dc at 10 mA/cm 2. Based on impedance analysis, plating at low frequency is dominated by the diffusion of cuprous or cupric ion while plating at high frequency is dominated by Cu 2ϩ ϩ e Ϫ ↔ Cu ϩ , resulting in the deposition improvement by high-frequency PC and PR. In the presence of PEGϩMPSϩCl, the uniformity improves with dc at 20 or 10 mA/cm 2. The trend in the baths with additives using PC and PR plating at high and low frequency is similar to that in the bath without additives. However, the improvement by high-frequency PC and PR is insignificant, since the plating is under adsorption control.

An accelerated learning algorithm (ABP-adaptive back propagation) is proposed for the supervised training of multilayer perceptron networks. The learning algorithm is inspired from the principle of "forced dynamics" for the total error functional. The algorithm updates the weights in the direction of steepest descent, but with a learning rate a specific function of the error and of the error gradient norm. This specific form of this function is chosen such as to accelerate convergence. Furthermore, ABP introduces no additional "tuning" parameters found in variants of the backpropagation algorithm. Simulation results indicate a superior convergence speed for analog problems only, as compared to other competing methods, as well as reduced sensitivity to algorithm step size parameter variations.

No abstract