An Improved Similarity-Based Clustering Algorithm for Multi-Database Mining.

Clustering algorithms for multi-database mining (MDM) rely on computing (n2-n)/2 pairwise similarities between n multiple databases to generate and evaluate m∈[1,(n2-n)/2] candidate clusterings in order to select the ideal partitioning that optimizes a predefined goodness measure. However, when these pairwise similarities are distributed around the mean value, the clustering algorithm becomes indecisive when choosing what database pairs are considered eligible to be grouped together. Consequently, a trivial result is produced by putting all the n databases in one cluster or by returning n singleton clusters. To tackle the latter problem, we propose a learning algorithm to reduce the fuzziness of the similarity matrix by minimizing a weighted binary entropy loss function via gradient descent and back-propagation. As a result, the learned model will improve the certainty of the clustering algorithm by correctly identifying the optimal database clusters. Additionally, in contrast to gradient-based clustering algorithms, which are sensitive to the choice of the learning rate and require more iterations to converge, we propose a learning-rate-free algorithm to assess the candidate clusterings generated on the fly in fewer upper-bounded iterations. To achieve our goal, we use coordinate descent (CD) and back-propagation to search for the optimal clustering of the n multiple database in a way that minimizes a convex clustering quality measure L(θ) in less than (n2-n)/2 iterations. By using a max-heap data structure within our CD algorithm, we optimally choose the largest weight variable θp,q(i) at each iteration i such that taking the partial derivative of L(θ) with respect to θp,q(i) allows us to attain the next steepest descent minimizing L(θ) without using a learning rate. Through a series of experiments on multiple database samples, we show that our algorithm outperforms the existing clustering algorithms for MDM.

Anaplastic thyroid carcinoma with chondrosarcomatous differentiation: a case report.

BACKGROUND: Anaplastic thyroid carcinoma (ATC) is a rare malignant tumor. In addition to the main ATC type with classical histopathological features, the other morphological types of ATC include paucicellular variant, angiomatoid, lymphoepithelioma-like, and small-cell variant. However, an ATC variant with a chondrosarcomatous component has not been reported to date. CASE PRESENTATION: Computed tomography imaging of a 63-year-old male with a 2-month history of a cervical mass revealed a 4.5-cm lesion with heterogeneous enhancement in the left thyroid lobe and two smooth and homogeneous nodules in the right thyroid lobe. The patient underwent total thyroidectomy and cervical lymph node resection. Histologically, the tumor boundary in the left lobe was clear, with a few mitotically active, spindle sarcoma-like tumor cells observed in some areas. Immunohistochemically, these spindle cells were positive for vimentin and negative for cytokeratin, paired box-8, epithelial membrane antigen, calcitonin, thyroglobulin, and thyroid transcription factor-1. In other areas, abundant cartilage matrix production and irregularly shaped lobules of cartilage, often separated by fibrous bands, were observed. The chondrocytes appeared mildly/moderately atypical and contained enlarged, hyperchromatic nucleoli. One of the two nodules in the right thyroid lobe had a clear boundary and comprised some bland spindle cells in a prominently collagenous stroma with clear boundaries. The other nodule in the right thyroid lobe was completely enclosed within a thin, fibrous capsule and exhibited normofollicular and microfollicular architecture. The patient received adjuvant radiotherapy after the surgery and was free of any local or regional recurrence or distant metastases at the 8-month follow-up evaluation. CONCLUSIONS: This unusual case of ATC with chondrosarcomatous differentiation is an important addition to the morphology spectrum of ATC types.

Cross-Layer Routing for a Mobility Support Protocol Based on Handover Mechanism in Cluster-Based Wireless Sensor Networks with Mobile Sink.

Wireless sensor networks with mobile collectors or sinks face some challenges regarding the data collection process and the continuous connectivity and delivering of data while the mobile sink is moving throughout the network. These challenges increase as the network grows. For this aim, we propose in this paper a cross-layer routing protocol which supports mobility for large-scale wireless sensor networks, which we name CLR-MSPH. We adapt CLR-MSPH for the hierarchical architecture of the network, and it performs on cluster-based wireless sensor networks where the network is organized in clusters. Our proposed protocol deals with the problem of handover data after the mobile sink leaves the radio range of cluster head without sending all data stored in the cluster head's buffer. We also introduce a mobility model for the mobile sink for a better data collection process. CLR-MSPH is considered as an extending implementation of BMAC protocol with handover mechanism (BMAC-H). In order to prove the efficiency of the proposed protocol, we compare CLR-MSPH to BMAC-H, where we adapted BMAC-H to perform in cluster-based wireless sensor networks. The simulation results show that CLR-MSPH performs better than BMAC-H in terms of packets reception rate, energy, and latency.