Aiming at the shortcomings of the traditional Density-Based Spatial Clustering of Applications with Noise -DBSCAN algorithm such as insignificant clustering effect and the choice of parameter combinations. This paper proposes an AD-DBSCAN algorithm with adaptive parameters, which makes the algorithm more difficult in the selection of the parameters. By establishing a DBSCAN algorithm model to adapt to finding the optimal distance threshold and the minimum number of neighbor points, the clustering is more accurate, and the noise point identified in the data is more accurate. Through the observation of the calculation model of the Calinski-Harabasz index, the evaluation index of the clustering algorithm, the selection of the optimal best distance threshold and the minimum number of neighborhood points, the accuracy of noise point recognition is improved by 5 times in the clustering algorithm, and the Calinski-Harabasz index improved by about 39.84%. The applicability of the algorithm in clustering the locations of urban road traffic accidents is verified.

INTRODUCTION: Parameters, such as left ventricular ejection fraction, peak strain dispersion, global longitudinal strain, etc. are influential and clinically interpretable for detection of cardiac disease, while manual detection requires laborious steps and expertise. In this study, we evaluated a video-based deep learning method that merely depends on echocardiographic videos from four apical chamber views of hypertensive cardiomyopathy detection. METHODS: One hundred eighty-five hypertensive cardiomyopathy (HTCM) patients and 112 healthy normal controls (N) were enrolled in this diagnostic study. We collected 297 de-identified subjects' echo videos for training and testing of an end-to-end video-based pipeline of snippet proposal, snippet feature extraction by a three-dimensional (3-D) convolutional neural network (CNN), a weakly-supervised temporally correlated feature ensemble, and a final classification module. The snippet proposal step requires a preliminarily trained end-systole and end-diastole timing detection model to produce snippets that begin at end-diastole, and involve contraction and dilatation for a complete cardiac cycle. A domain adversarial neural network was introduced to systematically address the appearance variability of echo videos in terms of noise, blur, transducer depth, contrast, etc. to improve the generalization of deep learning algorithms. In contrast to previous image-based cardiac disease detection architectures, video-based approaches integrate spatial and temporal information better with a more powerful 3D convolutional operator. RESULTS: Our proposed model achieved accuracy (ACC) of 92%, area under receiver operating characteristic (ROC) curve (AUC) of 0.90, sensitivity(SEN) of 97%, and specificity (SPE) of 84% with respect to subjects for hypertensive cardiomyopathy detection in the test data set, and outperformed the corresponding 3D CNN (vanilla I3D: ACC (0.90), AUC (0.89), SEN (0.94), and SPE (0.84)). On the whole, the video-based methods remarkably appeared superior to the image-based methods, while few evaluation metrics of image-based methods exhibited to be more compelling (sensitivity of 93% and negative predictive value of 100% for the image-based methods (ES/ED and random)). CONCLUSION: The results supported the possibility of using end-to-end video-based deep learning method for the automated diagnosis of hypertensive cardiomyopathy in the field of echocardiography to augment and assist clinicians. TRIAL REGISTRATION: Current Controlled Trials ChiCTR1900025325, Aug, 24, 2019. Retrospectively registered.

In recent years, the development of short-term traffic flow prediction technology has been the focus of many scholars. Although the existing traffic flow prediction methods perform well, they still fail to reach the level of accurate prediction. This is mainly because the model structure they adopted is simple, the factors considered are not enough, and the data processing methods they adopted are single. In this paper, a queue hybrid neural network (QHNN) model based on long short-term memory (LSTM) and gated recurrent unit (GRU), with weather weighted factor, is proposed to predict traffic flow. Queue hybrid neural network is proposed to extract the characteristics of traffic flow. The calculation formula of weather weighted factor is constructed to take more weather factors into consideration. The experimental results show that the method proposed in this paper is superior to the existing advanced models. The experimental process is more scientific because it is carried out in a step-by-step optimization way.

As WDM technology matures and multicast applications become increasingly popular, supporting multicast at the WDM layer becomes an important and yet challenging topic. In this paper, we study constrained multicast routing in WDM networks with sparse light splitting, i.e., where some switches are incapable of splitting light (or copying data in the optical domain). Specifically, we propose four WDM multicast routing algorithms, namely, Re-route-to Source, Re-route-to-Any, Member-First, and Member-Only. Given the network topology, multicast membership information, and light splitting capability of the switches, these algorithms construct a source-based multicast light-forest (consisting one or more multicast trees) for each multicast session. The performance of these algorithms are compared in terms of the average number of wavelengths used per forest (or multicast session), average number of branches involved (bandwidth) per forest as well as average number of hops encountered (delay) from a multicast source to a multicast member.

In high-speed SONET rings with point-to-point WDM links, the cost of SONET add-drop multiplexers (S-ADMs) can be dominantly high. However, by grooming traffic (i.e., multiplexing lower-rate streams) appropriately and using wavelength ADMs (WADMs), the number of S-ADMs can be dramatically reduced. In this paper, we propose optimal or near-optimal algorithms for traffic grooming and wavelength assignment to reduce both the number of wavelengths and the number of S-ADMs. The algorithms proposed are generic in that they can be applied to both unidirectional and bidirectional rings having an arbitrary number of nodes under both uniform and nonuniform (i.e., arbitrary) traffic with an arbitrary grooming factor. Some lower bounds on the number of wavelengths and S-ADMs required for a given traffic pattern are derived, and used to determine the optimality of the proposed algorithms. Our study shows that using the proposed algorithms, these lower bounds can he closely approached in most cases or even achieved in some cases. In addition, even when using a minimum number of wavelengths, the savings in S-ADMs due to traffic grooming (and the use of WADMs) are significant, especially for large networks.

As WDM technology matures, IP over WDM multicast will become a challenging new topic. Supporting multicast at the WDM layer provides additional advantages, but also raises many new issues that do not exist in IP multicast. For example, the limitation on the light splitting capability of switches is one major difficulty in WDM multicast, and in addition, the limitations on both the wavelength conversion capability and optical buffer space may affect multicast routing as well. In this paper, we focus on the IP over WDM multicast routing problem, i.e. how to construct multicast trees at the WDM layer based on IP multicast routing protocols. More specifically, we study how label switched paths for optical label switching can be set up for multicast traffic. We propose two approaches, one without modification of existing IP multicast routing protocols, and the other with modification of existing IP multicast routing protocols.

Supporting WDM multicasting in an IP over WDM network poses interesting problems because some WDM switches may be incapable of switching an incoming signal to more than one output interface. An approach to WDM multicasting based on wavelength-routing, which constructs a multicast forest for each multicast session so that multicast-incapable WDM switches do not need to multicast, was proposed and evaluated previously. Such an approach requires global knowledge of the WDM layer. In this paper, we study WDM multicasting in an IP over WDM network under the framework of multiprotocol label switching (MPLS) using optical burst/label switching (OBS/OLS). We propose a protocol which modifies a multicast tree constructed by distance vector multicast routing protocol (DVMRP) into a multicast forest based on the local information only.

We consider the problem of scheduling all-to-all personalized connections (AAPC) in WDM rings. Scheduling one connection for every source-destination pair in a network of limited connectivity provides a way to reduce routing control and guarantee throughput. For a given number of wavelengths K and a given number of transceivers per node T, we first determine the lower bound (LB) on the schedule length, which depends on both K and T. To achieve the LB, either the network bandwidth, the I/O capacity, or both should be fully utilized. This approach first constructs and then schedules circles, each of which is formed by up to four non-overlapping connections and can fully utilize the bandwidth of one wavelength. The proposed circle construction and scheduling algorithms can achieve the LB if K/spl les/T

We propose an on-line wavelength assignment algorithm for multi-fiber WDM networks, in which lightpaths are established and released dynamically. For a given number of fibers per link and number of wavelengths per fiber, the algorithm aims to minimize the blocking probability. It may also be used to reduce the number of wavelengths required for a given tolerable blocking probability. Simulation results show that our wavelength assignment algorithm performs better than other previously proposed algorithms (in the cases we studied). As the number of fibers per link increases, the benefit of having wavelength converters decreases dramatically, and the performance improvement of our algorithm over others increases. Our results also show that in case a preferred path is not available, rerouting along a node-disjoint backup path can significantly reduce the blocking probability.

Two properties of optical transmissions, namely, unidirectional propagation and predictable propagation delay, make it possible to pipeline packet transmissions in all-optical networks. In this paper, we study the problem of scheduling all-to-all personalized communication (AAPC) in unidirectional TDM/WDM rings with pipelined transmissions, which can achieve a much higher bandwidth utilization than non-pipelined transmissions. For a given number of wavelengths, K, and number of transmitter-receiver pairs per node, T, the theoretical lower bound (TLB) on the schedule lengths is derived and scheduling methods which can achieve near optimal results are proposed for three different cases, namely, T=K, 2/spl les/T

Selecting good speech units and building precise acoustic models on these units are basic problems for HMM speech recognition systems. In this paper, we review some distinguished phonetic features of the Chinese language and show how these features could be considered for getting a better solution of speech units selection and acoustic model building. The initials and finals are suggested to be the recognition units. Several experiments have been carried out to find those proper acoustic models which can represent the co-articulation among initials and finals more accurately and give a higher recognition rate. Results of recognition error rate under different acoustic models are given and some comparisons are made. Experiments show that not only the intra-syllable context-dependent acoustic models but also the inter-syllable acoustic models are necessary for reducing recognition error rate of words and continuous speech recognition. Experiments also show the tone-pattern information is important for accurate syllable recognition.