In this work an analysis of neurocontrollers is given. The purpose of this paper is the neurocontroler for attitude control: satellite rotations. The classification of neurocontroller architecture is provided. The pros and cons of different neurocontrollers are described. Two configuration of neural network – feedforward neural networks with mini-batch descent and modified Elman neural network, are investigated in this work to verify its ability to control the attitude of a satellite. The advantages and disadvantage of different predictive model neurorization systems are described. The class diagram for the simulating of satellite rotation for neural network learning is given. The proposed approach provides the architecture of the neural network and the weights among the layers in order to guarantee stability of the system. The accuracy was calculated. The AI module, after trained for different configurations of wheels, will get commands with desired 3D rotation speeds and control the wheels to achieve the desired rotation speeds.

The grid infrastructure has evolved as the integration and collaboration of multiple computer systems, networks, different databases and other network resources. The problem of scheduling in grid environment is an NP complete problem where conventional approaches like First Come First Serve (FCFS), Shortest Job First (SJF), Round Robin Scheduling algorithm (RR), Backfilling is not preferred because of the unexpectedly high computational cost and time in the worst case. Different algorithms, for example bio-inspired algorithms like Ant Colony Optimization (ACO), Artificial Bee Colony (ABC), Genetic Algorithm and Particle Swarm Optimization (PSO) are there which can be applied for solving NP complete problems. Among these algorithms, ACO is designed specifically to solve minimum cost problems and so it can be easily applied in grid environment to calculate the execution time of different jobs. Algorithms have different parameters and the performance of these algorithms extremely depends on the values of its parameters. In this paper, we have proposed a method to tune the parameters of ACO and discussed how parameter tuning affects the performance of ACO which in turn affects the performance of grid environment when applied for scheduling.

This research aims to study a development of a forecasting model to predict a monthly paddy price in Thailand with 2 datasets. Each of datasets is the univariate time series that is a monthly data, since Jan 1997 to Dec 2017. To generate a forecasting model, we present a forecasting model by using the Artificial Neural Network technique and determine its structure with Autoregressive form of the ARIMA model and Genetic Algorithm, it’s called AR-GA-ANN model. To generate the AR-GA-ANN model, we set 1 to 3 hidden layers for testing, determining the number of input nodes by an Autoregressive form of the ARIMA model and determine the number of neurons in hidden layer by Genetic Algorithm. Finally, we evaluate a performance of our AR-GA-ANN model by error measurement with Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE) and compare errors with the ARIMA model. The result found that all of AR-GA-ANN models have lower RMSE and MAPE than the ARIMA model and the AR-GA-ANN with 1 hidden layer has lowest RMSE and MAPE in both datasets.

Swarm intelligence algorithms are designed to mimic the natural behaviors of living organisms. The birds, animals and insects exhibit extraordinary problem solving behaviors and intelligence when living in colonies or groups. These unique behaviors form the basis for the design of the Metaheuristic which are helpful in solving several real-life combinatorial optimization problems. Monkey algorithm is developed based on the unique behaviors of monkeys such as mountain and tree climbing, jumping, watching and somersaulting. This paper reports for the first time the design and development of Multi-objective Monkey Algorithm (MoMA) and its use for the design of molecules with optimal drug-like properties. Finally, the performance of the proposed MoMA for Drug design (MoMADrug) is compared with the previously disclosed Multi-objective Genetic algorithm (MoGADdrug) for the design of drug-like molecules.

Most of action recognition methods allow achieving high action recognition accuracy, but only after processing the entire video sequence, however, for security issues, it is primordial to detect dangerous behavior occurrence as soon as possible allowing early warnings. In this paper, we present a human activity recognition method using 3D skeleton information, recovered by an RGB-D sensor by proposing a new descriptor modeling the dynamic relation between 3D locations of skeleton joints expressed in Euclidean distance and spherical coordinates between the normalized joints, A PCA dimension reduction is used to remove noisy information and enhance recognition accuracy while improving calculation and decision time. We also study the accuracy of the proposed descriptor calculated on limited few first frames and using limited skeleton joint number, to perform early action detection by exploring several classifiers. We test this approach on two datasets, MSR Daily Activity 3D and our own dataset called INDACT. Experimental evaluation shows that the proposed approach can robustly classify actions outperforming state-of-the-art methods and maintain good accuracy score even using limited frame number and reduced skeleton joints.

The aim of this paper is to detect multi-view faces using deep convolutional neural network (DCNN). Multi-view face detection is a challenging issue due to wide changes in appearance under different pose expression and illumination conditions. To address challenges, we designed a deep learning scheme with different network structures to enhance the multi view faces. More specifically, we design cascade architecture on convolutional neural networks (CNNs) which quickly reject non-face regions. Implementation, detection and retrieval of faces will be obtained with the help of direct visual matching technology. Further, a probabilistic calculation of resemblance among the images of face will be conducted on the basis of the Bayesian analysis for achieving detection of various faces. Experiment detects faces with ±90 degree out of plane rotations. Fine-tuned AlexNet is used to detect multi view faces. For this work, we extracted examples of training from AFLW (Annotated Facial Landmarks in the Wild) dataset that involve 21K images with 24K annotations of the face.