Search

Objectives: The landslides disaster is one of the most important factors for the construction of major infrastructure in the western China. How to effectively and reliably carry out wide-area landslide susceptibility prediction has always been a frontier difficulty in domestic and foreign studies. However, the present data-driven methods for prediction of landslide susceptibility in large-scale and complex scenarios still face two major issues: (1) The difference between various landslide inducing environments in a wide range of scenarios, would cause the difficulty to applying a single model to account for multiple landslide phenomenon; (2) small sample problem: Complex environmental tasks require models with large capacity and strong representative power, but there is a lack of sufficient landslide samples in practice. Methods: In response to the above problems, this paper takes Qijiang and Fuling District of Chongqing City, China as an example, proposes a local prediction strategy, and introduces the idea of meta-training an intermediate representation suited to be generalized, that can be adapted to the landslide sensitivity prediction task corresponding to the current local area, with only a very small number of samples in the local area, and number of iterations. Thus, the two issues mentioned can be well settled. Results: The proposed method is different from traditional methods such as support vector machines, multilayer perceptrons, and random forests, which require a large number of samples and gradient iterations to train the supervised model. Instead, only a small sample is required to fine-tune the intermediate model, which still improves the global accuracy by 1%-5%, the precision by 1%-3%, the F1-score by 0.5%-6%, and the recall rate is close to the highest level of other methods. Conclusions: The meta-learning paradigm enables few-shot adaptation of landslide susceptibility prediction model with superior statistical performance. [ABSTRACT FROM AUTHOR]

Easy clogging and short service life of mesh filters have posed great challenges in the irrigation water system. In this study, the pre- and post-pump mesh filters in series were combined to determine the interception of sediment and organic impurities at both pre- and post-pump filters. The indoor prototype tests were also carried out. The results showed that the distribution of the filter interception of sediment and organic impurities was related to the aperture of the mesh at pre- and postpump filters. The smaller the aperture of the post-pump filter was, the shorter the filtration time was. There was basically unchanged in the distribution of sawdust and sediment that was intercepted at pre- and post-pump filters. The smaller the aperture of the pre-pump filter was, the longer the filtration time was, and the smaller the percentage of sawdust intercepted by the filter. The percentage of sawdust intercepted by the pre-pump filter increased under the condition of the same screen aperture, with the increase of inlet water flow rate. It was recommended that the pre-pump filter with the screen aperture of 0.32 mm and the post-pump filter with the screen aperture of 0.20 mm were used for filtration when the content of organic impurities was high. Furthermore, the 0.32 and 0.20 mm mesh apertures of the pre- and post-pump filters were for filtration, when the mass ratio of organic impurities and sediment in the irrigation water source was less than 2. The 0.25-0.32 mm screen aperture of pre-pump filter prolonged the filtration time. There was a variation in the head loss with the filtration time, in order to clarify the influence of water conditions and screen aperture on the filtration time. Combined with the range analysis and multivariate analysis of variance (MANOVA), the initial and peak head loss were concentrated in the 2.43-5.87 and 13.92- 28.92 m, respectively, under different inlet flow conditions. The impurity ratio posed a smaller influence on the head loss. Additionally, the larger the inlet water flow rate was, the shorter the filtration time was. The empirical formula of head loss was established to fit the filtration time. The error between the fitting and the test was less than 7%, indicating the better suitable for the filtration time of tandem mesh filters under actual irrigation water. The best combination of the factor level for the orthogonal test was screened as the aperture D1（pre-pump is 0.32 mm, post-pump is 0.20 mm）, the sawdust to sediment mass ratio 1:1, sand content 0.12 g/L, and inlet flow 120 m3 /h. The influence of each factor on the filtration time was ranked in descending order: the impurity ratio, the screen aperture, the sand content, and the rate of inlet water flow. MLP neural network model was used to predict the filtration time. The error between the predicted and measured values was basically within 10%. The mean square error and the average relative error were 0.32% and 5.85%, respectively, suitable for the prediction of the filtration time under the complex water sources. The finding can also provide a strong reference to configuring the screen aperture of the pre- and post-pump filters in the tandem mesh filters for the irrigation projects. [ABSTRACT FROM AUTHOR]

In order to solve the problems of low and unstable prediction accuracy of the current yarn quality prediction model based on neural network, three prediction models were established, including RF (random forest) algorithm prediction model, MLP (multi-layer perceptron) neural network algorithm prediction model and LR (linear regression) algorithm prediction model. Experiments were conducted to compare the three models' prediction performance on the sensitivity with small samples in different data distribution, the sensitivity in different data dimensions and the sensitivity in different training sample sizes. The coefficient of determination and root mean square error were used to evaluate the prediction performance of each model. The experimental results showed that, compared with MLP algorithm and LR algorithm, in most cases, RF algorithm had higher prediction accuracy, better stability of prediction accuracy, better adaptability to small training sample size, and higher comprehensive prediction performance. [ABSTRACT FROM AUTHOR]

In order to solve the common problems of weak generalization ability and unstable prediction accuracy of neural network algorithm model on yarn quality prediction in the case of small samples in recent researches, a prediction model of random forest algorithm was established. A multi-layer perceptron (MLP) neural network algorithm model and a linear regression algorithm model were established as comparison models. The prediction experiments of yarn evenness and strength were carried out. The determination coefficient and prediction error were used to evaluate the prediction accuracy of the prediction model, and K fold cross validation was used to evaluate the generalization ability of the prediction models. The experimental results show that: in the case of small samples, compared with the MLP neural network algorithm model and the linear regression algorithm model, the random forest algorithm model has higher determination coefficient and smaller prediction error in the conventional prediction experiment and higher determination coefficients with smaller variation range in the K fold cross validation with different folding numbers. It is verified that the random forest algorithm model has good prediction accuracy, stability and generalization ability. [ABSTRACT FROM AUTHOR]

Copyright of Chinese Journal of Liquid Crystal & Displays is the property of Chinese Journal of Liquid Crystal & Displays and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

A numerical model is built using hydrodynamic analysis software and time domain analysis method according to the different mooring conditions of moored ships in port. Taking a 266 000 m³ LNG ship as an example, a database of the ship's six degrees of freedom motion is built. Based on the BP neural network, a prediction model for the movement of moored ships in the harbor is established. In the model design, the number of hidden layer nodes is 26, the number of input layer neurons is 12, the number of output layer neurons is 6, and the learning rate is 0.000 1. The training results are compared with the basic database, and the results show that the prediction results are in good agreement with the comparison results of the training sample database and the test sample database. Finally, based on the model test database, the prediction model results are compared with the test data of the physical model and the applicability of the method is verified, which provides a new idea for rapid prediction of the motions of moored ships in port. [ABSTRACT FROM AUTHOR]

axi destination prediction based on location services is an important aspect for reasonable planning of urban transportation. However, there exist problems of data sparsity and single feature of trajectory data, which affect the accuracy of destination prediction. To address the problem of data sparsity, the original taxi trajectory data (i. e. taxi speed, direction angle, and time) is combined with trajectory truncation method to determine the model input features, thus overcoming data sparsity. To address the problems of excessive parameters and overfitting caused by a large number of parameters of the multilayer perceptron (MLP), a parameter sharing mechanism based on convolution is proposed to solve parameter redundancy. Furthermore, an attention mechanism is used to allocate more computing resources to more important tasks, focusing on important information to improve the prediction performance of the model. Based on this, a taxi destination prediction method (CCMLP) that combines convolution, attention modules, and MLP is proposed, which achieves more accurate destination prediction while overcoming the structural deficiencies of MLP and convolution. The reliability of the CCMLP model is verified through different data and experiments. The experimental results show that selecting the trajectory features of the first m points and the last n points, as well as their corresponding direction angles, speeds, license plate numbers, and time features as the input features of the model can effectively improve the accuracy of destination prediction. The proposed CCMLP method has good feature learning ability. Compared with the MLP prediction model, the distance error is reduced by 10%, and compared with the ensemble learning algorithm, the distance error is reduced by 19.6%. The CCMLP model's generalization ability for different data distributions is verified by dividing the data set based on weekdays and weekends, with distance losses of 2.25 km and 2.23 km respectively. The effect of different trajectory completeness on destination prediction is verified based on the first ten points of the trajectory, with distance losses of 2.23 km, 1.80 km, 0.97 km, and 0.68 km for completeness rates of 40%, 60%, 80%, and complete trajectories respectively. [ABSTRACT FROM AUTHOR]

Copyright of China Sciencepaper is the property of China Sciencepaper and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Electric Power Automation Equipment / Dianli Zidonghua Shebei is the property of Electric Power Automation Equipment Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

System on wafer that incorporate advanced concepts such as dielets assembly and wafer integration are a new way to continue Moore's Law. Due to the uncertainty of the manufacturing yield and the splicing process in the wafer substrate network, the system on wafer has problems such as routing node failure or link failure. In order to improve the fault tolerance of the system, this paper proposed a fault-tolerant perceptron structure based on 2D-Mesh network on wafer. It used the primary and secondary perceptrons interchangeably in the Mesh ring to obtain the fault information of interconnected routing nodes, and used the broadcast mechanism to notify the global routing. The node could avoid obstacles and detours of data packets, and relieved the blocking of data packets that may occur in the routing process. Experimental simulations show that this structure has higher saturation injection rate than traditional fault-tolerant routing algorithms under various failure modes, and can effectively improve the fault tolerance of the system and reduce the performance impact caused by partial failures. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Beijing University of Technology is the property of Journal of Beijing University of Technology, Editorial Department and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Henan Agricultural Sciences is the property of Editorial Board of Journal of Henan Agricultural Sciences and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Aiming at the current hand pose estimation which doesn’t fully use 2D information to assistance 3D hand mesh reconstruction, this paper further introduced HOG feature of the RGB image based on the use of hand Mask and 2D heatmaps. This paper designed the framework of multi task learning CNN to extract the 2D information, and used the topology of the hand to fused the relevant information. In order to solve two problems: the consistency of hidden Embeddings dimensions in Transformer Encoder as well as requires heavy computational resources, the model introduced a new MLP (Multi-Layer Perception) module into transformer encoder to increase the length of Embeddings and decrease the dimensions of Embeddings, thereby the hand Mesh was completed in the way from coarse to fine. The new method proposed in this paper has achieves better results in two Frei-HAND and RHD datasets. [ABSTRACT FROM AUTHOR]

Northeast Rice is mainly grown in the plain areas of Heilongjiang, Jilin, and Liaoning provinces of China. The unique quality of Northeast rice can be attributed to the environmental advantages, including the fertile soil, sufficient sunshine, excellent water quality, long accumulated temperature, and large temperature difference between day and night. However, it is difficult to identify the Northeast rice in the market for the protection of regional special products. An accurate and rapid identification technology is of great significance to the Northeast rice origin. In this study, a total of 10 sampling areas were prepared in Heilongjiang, Jilin, and Liaoning provinces. 90 soil surface and rice samples were then collected. Inductively coupled plasma mass spectrometry (ICP-MS) was used to determine the content of 23 trace elements (such as Li, B, and Be) in 90 soil-crop seeds from the main rice-producing areas. The SPSS and SIMCA statistical analysis software was also used to analyze the distribution of trace elements in rice and soil from different producing areas. Correlation analysis showed that the contents of Mo and Zn in rice were positively correlated with the contents of Mo and Zn in soil. The analysis of variance showed that there was a consistent distribution of Ga, Pb, Sr, Zr, and Ba in rice from the three provinces, whereas, the rest 18 elements showed significant differences. Principal component analysis (PCA), partial least squares regression analysis (PLS-DA), orthogonal partial least squares regression analysis (OPLS-DA), fisher discriminant analysis (FDA), and multi-layer perceptron neural network (MLP-NN) were performed on the 18 elements with significant differences in rice. Furthermore, the cumulative variance of the first principal component and the second principal component was 46.39%, indicating only a little original variable information. There was no aggregate for the rice from the different provinces in two-dimensional space in the projection of the principal component score. By contrast, there was a small difference in rice element characteristics in the PLS-DA score chart, due to the geographical proximity. Meanwhile, confusion and cross phenomenon were found among rice samples from different producing areas. OPLS-DA, FDA, and MLP-NN were utilized to distinguish the rice from different producing areas. The OPLS-DA scores performed better to distinguish the rice from the Heilongjiang and Jilin provinces. There were a few overlaps in the samples between Jilin and Liaoning provinces, or between Heilongjiang and Liaoning provinces. The result of permutation test shows that the model established by orthogonal partial least squares regression analysis is reliable. In the FDA, the elements that were introduced into the Fisher discriminant model were B, Cr, Ni, Cu, Ge, Mo, and W in the order of stepwise discriminant analysis. The accuracy of the discriminant function was 93.8% for the original grouped cases, and 92.6% for the cross-validation of the rest. The multi-layer perceptron neural network was used to analyze 63 actual training samples, and 27 verification samples, with the group as the dependent variable, and 18 elements content as the covariable. The correct discrimination rate of training samples was 100%, and the comprehensive correct discrimination rate of the overall test group was 96.3%. Consequently, the different discrimination models, the content of trace elements in rice, and the characteristic elements can be expected to effectively distinguish the rice-producing areas of the three provinces in Northeast China. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Aiming at the problem of slow training speed and poor convergence of deep reinforcement learning method for the autopilot control parameters training, an intelligent training method that converts three-dimensional control parameters into one-dimensional design parameters is proposed with the three-loop autopilot pole placement method as the core. The intelligent control architecture of offline deep reinforcement learning training and online multi-layer perceptron neural network real-time calculation is constructed, which improves the efficiency and convergence of deep reinforcement learning algorithm and ensures the rapid online tuning of control parameters under the condition of large-scale flight state changes. Taking a typical reentry aircraft as an example, the deep reinforcement learning training and neural network deployment are accomplished. The simulation results show that the training efficiency of the simplified reinforcement learning action space is higher, and the tracking error of the controller to the control command is less than 1.2% by the proposed parameter rapid tuning method based on deep reinforcement learning. [ABSTRACT FROM AUTHOR]

In this study, the influence of the process parameter setting of the hot-air leaf moisturizer on the quality index of the exit leaf during the secondary leaf conditioning of threshing and redrying is studied, and the corresponding prediction model is established. A BP neural network prediction model is established based on the characteristics of the secondary leaf conditioning process data. The current popular neural network writing framework TensorFlow’s high-level API interface is called to construct the neural network structure. The activation function, optimizer, number of hidden layer neurons and other key parameters are gradually adjusted in the neural network structure to make the prediction result of the test set reach the best state. By inputting the parameters of the front steam nozzle pressure, front-end water flow rate, hot air temperature, return air temperature, feed blade temperature, and feed blade moisture combination, the two key tobacco leaf evaluation indicators, namely, outlet leaf moisture and temperature, are predicted. According to the mean square error, root mean square error, and average absolute error of the prediction results, it is concluded that when the number of neurons in the hidden layer is 7, the activation function selects ReLU, and the optimizer selects RMSprop, the effect is the best. [ABSTRACT FROM AUTHOR]

Apple diseases have pose a serious risk on the income of orchards in recent years. An accurate and rapid identification of apple diseases can be great benefit to better prevent and control diseases. Most effort has been made in the laboratory to train the identification model, due mainly to the limited condition for the deliberately infect apples in the real orchard. However, most models cannot fully meet the requirement of the disease detection in the large-scale production. In this study, a deep learning model (called CaTNet) was proposed to extract both the global and local information from the diseases of apple leaf. The image data of disease was collected from the apple orchards in the Jilin Province of China. A total of 16,464 images were obtained from the several publicly available datasets with the laboratory and natural environmental data collected from the field. Firstly, a model structure was constructed with both Transformer and convolutional neural network (CNN). Global and local information was extracted from the original images using the two branches. The strong generalization ability of the model was improved to learn a wider variety of features. Meanwhile, the global features were acquired to improve the resistance of the model to interference. Secondly, the Transformer block in the Transformer branch was optimized to make the structure simpler. In addition, a channel compression and expansion module was designed in the Transformer branch, in order to reduce the training cost of CaTNet for the less channel dimension of the input features. Afterwards, the multiple multilayer perceptrons were replaced by the grouped convolutional layers to further improve the computational speed of the model. Thirdly, the lightweight CNN branch was constructed with an inverse residual structure to fuse the point convolution of the expanded channels with the 3×3 convolution of the extracted information. The CNN branch was utilized to extract the local features of the image. As such, the model was more sensitive to the fine-grained features. Finally, the concat operation was implemented to fuse the different output of features from the two branches. After that, the CNN branch was selected to extract the local features from the global ones, whereas, the Transformer branch was extracted the global from the local. The multiple features to be cycled were also improved the generalization of the model. A comparison was made to clarify the effect of different down-sampling on the two-branch network. Specifically, an accuracy rate of 79.35%, 74.06% and 67.95% were obtained using pooling, 3×3 size convolution kernel, and 1×1 size convolution kernel for the down-sampling, respectively. The CaTNet model with two branches showed a computational speed of 0.108 2 s/Frame), which was faster than the various deep learning models, such as the EfficientNetV2 s (0.383 2 s/Frame) and PVT t (0.177 8 s/Frame). Consequently, the two-branch structure can be expected to accommodate more computation for the much higher computational speed. This finding can provide a design approach to build the deep learning models with the high generalization capability, particularly on the training with the high accuracy under only easily accessible data. [ABSTRACT FROM AUTHOR]

Copyright of Forest Research is the property of Forest Research Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Aiming at the knowledge overload in the process of online learning and the problems of data sparsity and cold start in the traditional recommendation algorithm, this paper proposed an improved learning resource recommendation algorithm based on MLP. The algorithm used the advantage of MLP in nonlinear data processing, converted the prediction mode of vector multiplication of learner characteristics and learning resource characteristics into the input mode of MLP, and improved the DN-CBR neural network recommendation model. In order to verify the effectiveness of the model, this paper built a dataset with the online learning platform data of iCourse as the sample. Compared with DN-CBR model, the improved model improves the normalized cumulative loss gain and hit rate by 1. 2 % and 3 % respectively on the dataset, which effectively improves the recommendation performance of the model. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Image & Graphics is the property of Editorial Office of Journal of Image & Graphics and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Due to the traditional end-to-end memory network model had insufficient feature representation ability, it couldn't represent the relationship between each memory well, which led to the low accuracy of location reasoning and path finding in the bAbI dataset. This paper proposed a new memory network combining density connectivity and multi-layer perceptron to solve this problem. This model used density connectivity and full connected layer to capture relationship features, which enhanced the capability of feature representation. The proposed model evaluated the accuracy of text reasoning using bAbI dataset. The experimental results show that compare with traditional me-mory network and the existing end-to-end memory network, the model can effectively improve the reasoning accuracy. [ABSTRACT FROM AUTHOR]

Copyright of Research of Environmental Sciences is the property of Research of Environmental Sciences Editorial Board and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

A quasi-ARX multilayer learning network prediction model was established and applied to the adaptive control of nonlinear systems. The kernel of the model is an improved neuro-fuzzy network: one part is a 3-layer nonlinear network with an off-line training self-associative network, the other part is a 3-layer neuro-fuzzy network adjusted online. Accordingly, the parameters were classified and the corresponding estimation algorithms were given. Then, the controller design scheme was proposed based on the advantages of the macrostructure of the model. Simulation analysis verifies the effectiveness of the proposed model. [ABSTRACT FROM AUTHOR]

In order to further excavate the hidden information between data, under the framework of multilayer concept factorization algorithm, this paper proposed a novel algorithm called dual-graph regularized multilayer concept factorization algorithm, which encoded the geometric structure information of data and feature spaces by constructing two Laplacian regularize term in each layer factorization, respectively. By this way, the proposed method could learn features in a hierarchical manner, and thus provided a better chance for learning meaningful features from the complex data. Moreover, it developed the iterative updating optimization scheme for DHCF, and also provided the convergence proof of the optimization scheme. Experimental results on TDT2 document datasets, PIE and COIL20 image datasets demonstrate the effectiveness of this proposed method. [ABSTRACT FROM AUTHOR]

Due to the difficulty of estimating probability function and the high space complexity of the existing probabilistic neural network (PNN), an improved PNN model is presented by introducing the mechanism of back-propagation (BP). The improved model inherits the principle and structure of PNN, and meanwhile applies the BP algorithm of multilayer perceptron (MLP) to train probability function and parameters. The above two aspects help overcome the PNN's shortcomings. Three numerical experiments have been designed to verify the improved model, and their results indicate that the new model has strong capacity to identify the importance of input indicators and own high accuracy of classification. In conclusion, the BP-PNN model is a new pattern classification method with widespread applicability and can support management decision. [ABSTRACT FROM AUTHOR]

A self-supervised multilayer perceptron online learning algorithm was proposed to improve the adaptability and real-time performance for unmanned ground vehicle unstructured road recognition. The road recognition results were used to update the training data set characteristic vector, and an evaluation function was created to trigger classifier retraining, as a result, the current classifier can recognize the road surface efficiently. Also, in the algorithm the processing operations such as the road surface image data sampling, classifier training, training data set updating and classifier recognition were calculated in their own threading. The structure can take advantage of faster classification calculation character of multilayer perceptron, and overcome its problem of time consuming training process. The real vehicle road recognition tests show that the proposed algorithm has a better adaptability and can meet the real-time requirements of unmanned ground vehicle unstructured road navigation. [ABSTRACT FROM AUTHOR]

In order to overcome the shortcomings of feedforward neural network's slow convergence, involving the local optimal and depending on the initial weights, this paper proposed a new method to train feedforward neural networks (FNNs) parameters based on the iterative chaotic map with infinite collapses particle swarm optimization (ICMICPSO) algorithm. This algorithm made full use of the information of BP's error back propagation and gradient. It used ICMICPS as the global optimizer to adjust the neural networks' weights and thresholds, when network parameters converge around global optimum. And it used gradient information as a local optimizer to accelerate the modification at a local scale. Compared with other algorithms, results show that the performance of the ICMICPSO-BPNN method is superior to the contrast methods in training and generalization ability. [ABSTRACT FROM AUTHOR]