A Gyroless Algorithm with Multi-Hypothesis Initialization for Projectile Navigation

Projectiles are subjected to a high acceleration shock at launch (20,000 g and higher) and can spin very fast. Thus, the components of onboard navigation units must therefore withstand such constraints in addition to being inexpensive. This makes only a few inertial sensors suitable for projectiles navigation. Particularly, rate gyroscopes which are gun-hardened and have an appropriate operating range are not widely available. On the other hand, magneto-resistive sensors are inexpensive and can satisfy both gun-hardening and operating range requirements, making them an alternative for angular estimation in guided projectiles. This paper presents a gyroless navigation algorithm for projectiles. The lack of gyroscope is handled by the usage of attitude kinematics computed over past attitude estimates of the filter, coupled with a measurement model based on magnetometer and GPS observations of the attitude. The observability of the attitude when considering non-calibrated magnetometers and its dependency on the initialization is addressed. Then, to cope with the initialization dependency of the filter, we proposed a multi-hypothesis initialization algorithm. In terms of performance, the algorithm is shown to provide a high-rate navigation solution with an interesting performance.

Analyzing and Accelerating the Bottlenecks of Training Deep SNNs With Backpropagation

Spiking neural networks (SNNs) with the event-driven manner of transmitting spikes consume ultra-low power on neuromorphic chips. However, training deep SNNs is still challenging compared to convolutional neural networks (CNNs). The SNN training algorithms have not achieved the same performance as CNNs. In this letter, we aim to understand the intrinsic limitations of SNN training to design better algorithms. First, the pros and cons of typical SNN training algorithms are analyzed. Then it is found that the spatiotemporal backpropagation algorithm (STBP) has potential in training deep SNNs due to its simplicity and fast convergence. Later, the main bottlenecks of the STBP algorithm are analyzed, and three conditions for training deep SNNs with the STBP algorithm are derived. By analyzing the connection between CNNs and SNNs, we propose a weight initialization algorithm to satisfy the three conditions. Moreover, we propose an error minimization method and a modified loss function to further improve the training performance. Experimental results show that the proposed method achieves 91.53% accuracy on the CIFAR10 data set with 1% accuracy increase over the STBP algorithm and decreases the training epochs on the MNIST data set to 15 epochs (over 13 times speed-up compared to the STBP algorithm). The proposed method also decreases classification latency by over 25 times compared to the CNN-SNN conversion algorithms. In addition, the proposed method works robustly for very deep SNNs, while the STBP algorithm fails in a 19-layer SNN.

A Fast Algorithm to Initialize Cluster Centroids in Fuzzy Clustering Applications

The goal of partitioning clustering analysis is to divide a dataset into a predetermined number of homogeneous clusters. The quality of final clusters from a prototype-based partitioning algorithm is highly affected by the initially chosen centroids. In this paper, we propose the InoFrep, a novel data-dependent initialization algorithm for improving computational efficiency and robustness in prototype-based hard and fuzzy clustering. The InoFrep is a single-pass algorithm using the frequency polygon data of the feature with the highest peaks count in a dataset. By using the Fuzzy C-means (FCM) clustering algorithm, we empirically compare the performance of the InoFrep on one synthetic and six real datasets to those of two common initialization methods: Random sampling of data points and K-means++. Our results show that the InoFrep algorithm significantly reduces the number of iterations and the computing time required by the FCM algorithm. Additionally, it can be applied to multidimensional large datasets because of its shorter initialization time and independence from dimensionality due to working with only one feature with the highest number of peaks.

SELF-ORGANIZING MAP WITH NGUYEN-WIDROW INITIALIZATION ALGORITHM FOR GROUNDWATER VULNERABILITY ASSESSMENT

Assessment of groundwater vulnerability to contamination plays a vital role in the utilization and protection of groundwater resource. In this study, a vulnerability map for Boracay Island, Philippines was developed using a modified self-organizing map algorithm to determine groundwater vulnerability in light of massive tourism developments in the island. Self-organizing map using the Nguyen-Widrow initialization algorithm was used to cluster DRASTIC data which were pre-processed using data cleaning normalization schemes. The vulnerability map developed showed that groundwater resource in the island is susceptible to contamination as confirmed by groundwater quality analysis. The result of the study demonstrates the effectiveness of the improved SOM algorithm as a tool for assessment of groundwater vulnerability and is comparable with the traditional DRASTIC method. The developed methodology allows grouping of datasets into clusters that represent the level of vulnerability to contamination of the groundwater. Further, this approach can be applied to other islands to ensure the balance between tourism developments and ecological integrity of the scarce groundwater resource.

Improved Initialization of the EM Algorithm for Mixture Model Parameter Estimation

A commonly used tool for estimating the parameters of a mixture model is the Expectation–Maximization (EM) algorithm, which is an iterative procedure that can serve as a maximum-likelihood estimator. The EM algorithm has well-documented drawbacks, such as the need for good initial values and the possibility of being trapped in local optima. Nevertheless, because of its appealing properties, EM plays an important role in estimating the parameters of mixture models. To overcome these initialization problems with EM, in this paper, we propose the Rough-Enhanced-Bayes mixture estimation (REBMIX) algorithm as a more effective initialization algorithm. Three different strategies are derived for dealing with the unknown number of components in the mixture model. These strategies are thoroughly tested on artificial datasets, density–estimation datasets and image–segmentation problems and compared with state-of-the-art initialization methods for the EM. Our proposal shows promising results in terms of clustering and density-estimation performance as well as in terms of computational efficiency. All the improvements are implemented in the rebmix R package.

Implementation of Parallelized K-means and K-Medoids++ Clustering Algorithms on Hadoop Map Reduce Framework

The electronic information from online newspapers, journals, conference proceedings website pages and emails are growing rapidly which are generating huge amount of data. Data grouping has been gotten impressive consideration in numerous applications. The size of data is raised exponentially due to the advancement of innovation and development, makes clustering of vast size of information, a challenging issue. With the end goal to manage the issue, numerous scientists endeavor to outline productive parallel clustering representations to be needed in algorithms of hadoop. In this paper, we show the implementation of parallelized K-Means and parallelized K-Medoids algorithms for clustering an large data objects file based on MapReduce for grouping huge information. The proposed algorithms combines initialization algorithm with Map Reduce framework to reduce the number of iterations and it can scale well with the commodity hardware as the efficient process for large dataset processing. The outcome of this paper shows the implementation of each algorithms.