Search

Iterative procedures for optimizing properties of molecular models often converge slowly owing to the computational cost of accurately representing features of interest. Here, we introduce a preconditioning scheme that allows one to use a less expensive model to guide exploration of the energy landscape of a more expensive model and thus speed the discovery of locally stable states of the latter. We illustrate our approach in the contexts of energy minimization and the string method for finding transition pathways. The relation of the method to other multilevel simulation techniques and possible extensions are discussed.

The adaptive coupling of the laser beam from space to single-mode fiber plays an important role in free space optical communication. A typical adaptive algorithm is stochastic parallel gradient descent algorithm (SPGD), which measures the performance index value to estimate the gradient value to maximize the coupling efficiency. It is conceivable that the presence of performance index measurement noise will have a great influence on the convergence performance of the algorithm. We propose an improved Kalman stochastic parallel gradient descent algorithm (KSPGD). Specifically, considering the influence of measurement noise on gradient estimation, we introduce the gradient prediction model in the iterative optimization process and then use the Kalman filter to estimate the gradient of the current iteration point. Kalman filter algorithm and optimization algorithm are integrated together. Simulation and experimental results show that the KSPGD algorithm can better restrain the influence of measurement noise on the convergence performance of the algorithm than the SPGD algorithm. [ABSTRACT FROM AUTHOR]

The emergence of resource conflicts and overload control problems during the playback of smart TV terminals has brought many obstacles to the operation of smart TV terminals, which has seriously affected the user experience of smart TV terminal users. In this regard, the adaptive media playback algorithm is optimized for smart TV terminals. This method performs dynamic priority preemptive scheduling on exclusive resources according to resource characteristics and application priorities to optimize resource allocation and improve media playback. The feedback control algorithm is used to perform QoS scheduling on shared resources until QoS proportional fairness is achieved, and QoS proportional compression is used to eliminate resource overload. Finally, a DASH server on Apache and implements an analog DASH client using Python are built. In order to verify the performance of the algorithm, the research results show that the adaptive media playback algorithm has the overload control capability, which only solves the resource conflict and improves the response performance under heavy load of the system, and the algorithm consumes only 4.5% of the overall system QoS. Compared with the existing methods, it is about 30% lower, which is more suitable for resource scheduling of smart TV terminals. The research in this paper shows that the adoption of QoS scheduling mechanism contributes to the optimization of media playback resources and allocation ratio, thus making the playback process of smart TV terminals, which provides a reference for the optimization of smart TV terminal playback. [ABSTRACT FROM AUTHOR]

In free space optical (FSO) communication, atmospheric turbulence will introduce aberration to the laser wavefront, which will reduce the coupling efficiency of space light to optical fiber. The system based on blind optimization principle can realize wavefront aberration compensation by using actuators such as fast steering mirror (FSM) and deformable mirror (DM), combined with blind optimization control algorithms. Stochastic parallel gradient descent (SPGD) algorithm is widely used in this kind of system. The disadvantages of SPGD algorithm are possibly local convergence and slow convergence. In order to improve the convergence performance of the SPGD algorithm, we proposed improved SPGD algorithms called Adagrad & Staged SPGD algorithm (AS-SPGD). Specifically, to accelerate SPGD algorithm, the Adagrad SPGD algorithm (A-SPGD) integrated Adagrad's adaptive gain strategy is proposed firstly in wavefront tip–tilt aberration compensation using FSM. Then, we consider compensating the wavefront high-order aberration using DM. By adopting the strategy of using SPGD algorithm with different settings in different stages, the Staged SPGD algorithm (S-SPGD) is proposed to improve the local convergence of the traditional SPGD algorithm. The simulation results show that the Adagrad SPGD algorithm converges faster than the traditional SPGD algorithm in the compensation of tip–tilt aberrations of different amplitude; the Staged SPGD algorithm has better global convergence ability than the traditional SPGD algorithm in high-order aberration compensation. • The blind optimization control is used to realize wavefront aberration compensation. • The Adagrad SPGD algorithm is proposed to improve the convergence speed. • The Staged SPGD algorithm is proposed to improve the local convergence. • Simulation results verify the modified SPGD methods have better performance. [ABSTRACT FROM AUTHOR]

• We propose a simple but efficient pre-processing algorithm capable of extracting translation positions of the structure light directly from raw images of Incoherent pattern-illumination Fourier ptychography (IFP), termed translation position extracting (TPE). • TPE removes the requirement of IFP for prior knowledge of translation positions, thus further relaxing the practical conditions. Both simulation and experiment are performed to demonstrate its effectiveness. • The calculation cost of the TPE algorithm is similar to the cross-correlation, which is far lower than the existing translation extraction algorithm based on iterative optimization. Incoherent Pattern-illuminated Fourier ptychography (IFP) is a newly developed super-resolution method, where accurate prior knowledge of translation positions is essential for image reconstruction. To release this limitation, we propose a pre-processing algorithm capable of extracting translation positions of the structure light directly from raw images of IFP, termed translation position extracting (TPE). TPE mainly involves two steps. First, the speckle parts mixed in the acquired raw images, in which the illumination motion is encoded, are isolated by intensity averaging and division. Then the cross-correlations of the speckle dataset are computed to determine the shift positions. TPE-IFP improves the previous IFP by removal of the requirement for prior knowledge of translation positions. Its effectiveness is demonstrated by obtaining high-quality super-resolution images in absence of location information in both simulations and experiments. By further relaxing the practical conditions, the proposed TPE may accelerate the applications of IFP. What's more, as a pre-processing approach, TPE might also contribute to the estimation of pattern positions for the similar speckle-based imaging. [ABSTRACT FROM AUTHOR]