Molecular insight into the GaP(110)-water interface using machine learning accelerated molecular dynamics

GaP has been shown to be a promising photoelectrocatalyst for selective CO_(2)reduction to methanol.Due to the relevance of the interface structure to important processes such as electron/proton transfer,a detailed understanding of the GaP(110)-water interfacial structure is of great importance.Ab initio molecular dynamics(AIMD)can be used for obtaining the microscopic information of the interfacial structure.However,the GaP(110)-water interface cannot converge to an equilibrated structure at the time scale of the AIMD simulation.In this work,we perform the machine learning accelerated molecular dynamics(MLMD)to overcome the difficulty of insufficient sampling by AIMD.With the help of MLMD,we unravel the microscopic information of the structure of the GaP(110)-water interface,and obtain a deeper understanding of the mechanisms of proton transfer at the GaP(110)-water interface,which will pave the way for gaining valuable insights into photoelectrocatalytic mechanisms and improving the performance of photoelectrochemical cells.

On the learning dynamics of two-layer quadratic neural networks for understanding deep learning

Deep learning performs as a powerful paradigm in many real-world applications;however,its mechanism remains much of a mystery.To gain insights about nonlinear hierarchical deep networks,we theoretically describe the coupled nonlinear learning dynamic of the two-layer neural network with quadratic activations,extending existing results from the linear case.The quadratic activation,although rarely used in practice,shares convexity with the widely used ReLU activation,thus producing similar dynamics.In this work,we focus on the case of a canonical regression problem under the standard normal distribution and use a coupled dynamical system to mimic the gradient descent method in the sense of a continuous-time limit,then use the high order moment tensor of the normal distribution to simplify these ordinary differential equations.The simplified system yields unexpected fixed points.The existence of these non-global-optimal stable points leads to the existence of saddle points in the loss surface of the quadratic networks.Our analysis shows there are conserved quantities during the training of the quadratic networks.Such quantities might result in a failed learning process if the network is initialized improperly.Finally,We illustrate the comparison between the numerical learning curves and the theoretical one,which reveals the two alternately appearing stages of the learning process.

A hybrid CNN-LSTM model for diagnosing rice nutrient levels at the rice panicle initiation stage

Nitrogen(N)and potassium(K)are two key mineral nutrient elements involved in rice growth.Accurate diagnosis of N and K status is very important for the rational application of fertilizers at a specific rice growth stage.Therefore,we propose a hybrid model for diagnosing rice nutrient levels at the early panicle initiation stage(EPIS),which combines a convolutional neural network(CNN)with an attention mechanism and a long short-term memory network(LSTM).The model was validated on a large set of sequential images collected by an unmanned aerial vehicle(UAV)from rice canopies at different growth stages during a two-year experiment.Compared with VGG16,AlexNet,GoogleNet,DenseNet,and inceptionV3,ResNet101 combined with LSTM obtained the highest average accuracy of 83.81%on the dataset of Huanghuazhan(HHZ,an indica cultivar).When tested on the datasets of HHZ and Xiushui 134(XS134,a japonica rice variety)in 2021,the ResNet101-LSTM model enhanced with the squeeze-and-excitation(SE)block achieved the highest accuracies of 85.38 and 88.38%,respectively.Through the cross-dataset method,the average accuracies on the HHZ and XS134 datasets tested in 2022 were 81.25 and 82.50%,respectively,showing a good generalization.Our proposed model works with the dynamic information of different rice growth stages and can efficiently diagnose different rice nutrient status levels at EPIS,which are helpful for making practical decisions regarding rational fertilization treatments at the panicle initiation stage.

Application of improved BPNN in image restoration-learning coefficient

A new method of artificial intelligence based on a new improved back propagation neural network （BPNN） algorithm is partially applied in the problem of image restoration. In order to over- come the inherited issues in conventional back propagation algorithm i.e. slow convergence rate, longer training time, hard to achieve global minima etc. , different methods have been used including the introduction of dynamic learning rate and dynamic momentum coefficient etc. With the passage of time different techniques has been used to improve the dynamicity of these coefficients. The meth- od applied in this paper improves the effect of learning coefficient η by using a new way to modify the value dynamically during learning process. The experimental results show that this helps in im- proving the efficiency overall both in visual effect and quality analysis.

Research of Dynamic Competitive Learning in Neural Networks

Introduce a method of generation of new units within a cluster and aalgorithm of generating new clusters. The model automatically builds up its dynamically growinginternal representation structure during the learning process. Comparing model with other typicalclassification algorithm such as the Kohonen's self-organizing map, the model realizes a multilevelclassification of the input pattern with an optional accuracy and gives a strong support possibilityfor the parallel computational main processor. The idea is suitable for the high-level storage ofcomplex datas structures for object recognition.

Call for papers Journal of Control Theory and Applications Special issue on Approximate dynamic programming and reinforcement learning

Approximate dynamic programming (ADP) is a general and effective approach for solving optimal control and estimation problems by adapting to uncertain and nonconvex environments over time.

Hydraulic Circuit Design and Dynamic Learning Using Case-based Reasoning

This paper describes the design and implementation of a hydraulic circuit design system using case-based reasoning (CBR) paradigm from AI community The domain of hydraulic circuit design and case-based reasoning are briefly reviewed Then a proposed methodology in compuer-aided circuit design and dynamic leaning with the use of CBR is described Finally an application example is selected to illustrate the ussfulness of applying CBR in hydraulic circuit design with leaming.

A multiscale differential-algebraic neural network-based method for learning dynamical systems

The objective of dynamical system learning tasks is to forecast the future behavior of a system by leveraging observed data.However,such systems can sometimes exhibit rigidity due to significant variations in component parameters or the presence of slow and fast variables,leading to challenges in learning.To overcome this limitation,we propose a multiscale differential-algebraic neural network(MDANN)method that utilizes Lagrangian mechanics and incorporates multiscale information for dynamical system learning.The MDANN method consists of two main components:the Lagrangian mechanics module and the multiscale module.The Lagrangian mechanics module embeds the system in Cartesian coordinates,adopts a differential-algebraic equation format,and uses Lagrange multipliers to impose constraints explicitly,simplifying the learning problem.The multiscale module converts high-frequency components into low-frequency components using radial scaling to learn subprocesses with large differences in velocity.Experimental results demonstrate that the proposed MDANN method effectively improves the learning of dynamical systems under rigid conditions.

Feature Selection and Feature Learning for High-dimensional Batch Reinforcement Learning: A Survey

Tremendous amount of data are being generated and saved in many complex engineering and social systems every day.It is significant and feasible to utilize the big data to make better decisions by machine learning techniques. In this paper, we focus on batch reinforcement learning(RL) algorithms for discounted Markov decision processes(MDPs) with large discrete or continuous state spaces, aiming to learn the best possible policy given a fixed amount of training data. The batch RL algorithms with handcrafted feature representations work well for low-dimensional MDPs. However, for many real-world RL tasks which often involve high-dimensional state spaces, it is difficult and even infeasible to use feature engineering methods to design features for value function approximation. To cope with high-dimensional RL problems, the desire to obtain data-driven features has led to a lot of works in incorporating feature selection and feature learning into traditional batch RL algorithms. In this paper, we provide a comprehensive survey on automatic feature selection and unsupervised feature learning for high-dimensional batch RL. Moreover, we present recent theoretical developments on applying statistical learning to establish finite-sample error bounds for batch RL algorithms based on weighted Lpnorms. Finally, we derive some future directions in the research of RL algorithms, theories and applications.

Dynamic neighborhood genetic learning particle swarm optimization for high-power-density electric propulsion motor

To maximize the power density of the electric propulsion motor in aerospace application,this paper proposes a novel Dynamic Neighborhood Genetic Learning Particle Swarm Optimization(DNGL-PSO)for the motor design,which can deal with the insufficient population diversity and non-global optimal solution issues.The DNGL-PSO framework is composed of the dynamic neighborhood module and the particle update module.To improve the population diversity,the dynamic neighborhood strategy is first proposed,which combines the local neighborhood exemplar generation mechanism and the shuffling mechanism.The local neighborhood exemplar generation mechanism enlarges the search range of the algorithm in the solution space,thus obtaining highquality exemplars.Meanwhile,when the global optimal solution cannot update its fitness value,the shuffling mechanism module is triggered to dynamically change the local neighborhood members.The roulette wheel selection operator is introduced into the shuffling mechanism to ensure that particles with larger fitness value are selected with a higher probability and remain in the local neighborhood.Then,the global learning based particle update approach is proposed,which can achieve a good balance between the expansion of the search range in the early stage and the acceleration of local convergence in the later stage.Finally,the optimization design of the electric propulsion motor is conducted to verify the effectiveness of the proposed DNGL-PSO.The simulation results show that the proposed DNGL-PSO has excellent adaptability,optimization efficiency and global optimization capability,while the optimized electric propulsion motor has a high power density of 5.207 kW/kg with the efficiency of 96.12%.

Consumers can learn and can forget-Modeling the dynamic decision procedure when watching TV

Facing the challenge of attracting consumers and winning market share under the pro-liferation of TV stations and channels,the traditional TV stations often make some mar-keting strategies.However,how to evaluate the effectiveness of different strategies and select the best one is a key issue.This study proposes to resolve this problem.We develop an innovative structural model to simulate the dynamic choices consumers make under two interactive behaviors:learning and forgetting.Learning behavior refers to updating programme quality assessment by using experience,while forgetting behavior prevents the use of previous experience.The Bayesian rules are employed to model learning behavior,and they are extended by incorporating an exponential decay function to mea-sure the effect of forgetting behavior.The structural model is tested and validated by using Hong Kong television viewing data.The empirical results show that when modeling consumer choice decisions,considering learning and forgetting behavior significantly improves the performance of the model in regard to rating prediction and marketing strategy evaluation.Five cases are simulated to show how the model is used to evaluate marketing strategies.Managerial implications are then discussed to guide the decision-making of traditional TV broadcasters and advertisers.

Fuzzy Genetic Sharing for Dynamic Optimization

Recently,genetic algorithms(GAs) have been applied to multi-modal dynamic optimization(MDO).In this kind of optimization,an algorithm is required not only to find the multiple optimal solutions but also to locate a dynamically changing optimum.Our fuzzy genetic sharing(FGS) approach is based on a novel genetic algorithm with dynamic niche sharing(GADNS).FGS finds the optimal solutions,while maintaining the diversity of the population.For this,FGS uses several strategies.First,an unsupervised fuzzy clustering method is used to track multiple optima and perform GADNS.Second,a modified tournament selection is used to control selection pressure.Third,a novel mutation with an adaptive mutation rate is used to locate unexplored search areas.The effectiveness of FGS in dynamic environments is demonstrated using the generalized dynamic benchmark generator(GDBG).