Recent deep-learning approaches have shown that Frequency Transformation (FT) blocks can significantly improve spectrogram-based single-source separation models by capturing frequency patterns. The goal of this paper is to extend the FT block to fit the multi-source task. We propose the Latent Source Attentive Frequency Transformation (LaSAFT) block to capture source-dependent frequency patterns. We also propose the Gated Point-wise Convolutional Modulation (GPoCM), an extension of Feature-wise Linear Modulation (FiLM), to modulate internal features. By employing these two novel methods, we extend the Conditioned-U-Net (CUNet) for multi-source separation, and the experimental results indicate that our LaSAFT and GPoCM can improve the CUNet's performance, achieving state-of-the-art SDR performance on several MUSDB18 source separation tasks.

No abstract

Hadoop distributed file system (HDFS) is designed to store huge data set reliably, has been widely used for processing massive-scale data in parallel. In HDFS, the data locality problem is one of critical problem that causes the performance decrement of a file system. To solve the data locality problem, we propose an efficient data replication scheme based on access count prediction in a Hadoop framework. By the previous data access count, the existing data replication scheme predicts the next access count of data files using Lagrange's interpolation. Then, the proposed data replication scheme determines the replication factor with the predicted data access count, whether it generates a new replica or it uses the loaded data as cache selectively. Finally, the proposed scheme provides improvement of data locality. By performance evaluation, proposed efficient data replication scheme is compared with default data replication setting of Hadoop that shows proposed scheme reduces averagely 8.9% of the task completion time in the map phase. Regarding the data locality, proposed scheme provides the increase of node locality by 6.6% and the decrease of rack and rack-off locality by 38.9% and 56.5%.

The key issue in mobile computing environments (MCEs) is how to relieve communication congestion and provide accurate information through location-based services (LBSs). The nearest surrounder (NS) query, used to find all visible objects around a given location, is a type of spatial query that suggests broad application base in LBS domain. However, because existing works for NS query only take into account static query points, the application of the NS query is limited to various LBSs in MCEs requiring frequent location updates. Motivated by this limitation, this paper introduces the continuous nearest surrounder (CNS) query, which uses a decentralized system framework to continuously maintain updated query results in MCEs. In this framework, the LBS server executes an initial NS query to prepare a region, termed non-provoked polygon (NPP), defines a set of visible objects that cannot be changed. Conversely, a client caches the NPP and does not update request unless it leaves its NPP. We performed extensive experiments using synthetic and real datasets with various data cardinality, and query mobility to validate the accurate performance of the proposed strategy. The results show that the CNS algorithm outperforms NS, in terms of computation and communication costs as well as scalability.ADV. PROCESSING TECHNOLOGIES AND APPLICATIONS FOR MOBILE COMM. SYSTEMS 771 the location of the mobile device which moves continuously and randomly, NS results become obsolete whenever a user moves to a new location. Figure 1 summarizes the concept of the NS query for a user moving in an MCE. Let Q be a query route Q D ¹Q 1 , Q 2 , : : : , Q m º in a time interval OEt s , t e , where Q i D .q i ,1 , q i ,2 / in R 2 . Initially, the result of the NS query at Q 1 is NS

No abstract