Recently, several parallel frameworks have emerged to utilize the increasing computational capacity of multiprocessors. Parallel tasks are distinguished from traditional sequential tasks in that the subtasks contained in a single parallel task can simultaneously execute on multiple processors. In this study, we consider the scheduling problem of minimizing the number of processors on which the parallel real-time tasks feasibly run. In particular, we focus on scheduling sporadic parallel real-time tasks, in which precedence constraints between subtasks of each parallel task are expressed using a directed acyclic graph (DAG). To address the problem, we formulate an optimization problem that aims to minimize the maximum processing capacity for executing the given tasks. We then suggest a polynomial solution consisting of three steps: (1) transform each parallel real-time task into a series of multithreaded segments, while respecting the precedence constraints of the DAG; (2) selectively extend the segment lengths; and (3) interpret the problem as a flow network to balance the flows on the terminal edges. We also provide the schedulability bound of the proposed solution: it has a capacity augmentation bound of 2. Our experimental results show that the proposed approach yields higher performance than one developed in a recent study.

To support ever-chainging user needs such as large storage volumes, web search, and high-performance computing, numerous companies have expanded their systems to cloud computing servers. Cloud environment systems generally consume large amounts of electrical power, leading to tremendously high operational costs. In addition, they require computing infrastructures to run various real-time applications such as financial analysis, cloud gaming, and web-based real-time services. To represent performance guarantees, the negotiated agreements in real-time computing, expressed as deadline (or latency), can be specified by service level agreements of cloud services between users and cloud server providers. Thus, a number of research works have started focusing on reducing the energy consumption and simultaneously satisfying the temporal constraint in a cloud environment. Although we previously proposed an optimal real-time scheduling algorithm for multiprocessors, it is difficult to use it for cloud environments handling a large number of cloud services because of the high computational complexity of $\Omega (N^{3}logN)$ , where $N$ is the number of tasks. Thus, we introduce a real-time task scheduling algorithm for cloud computing servers, which alleviates the computational complexity of $O(N^{2})$ from the complexity of the previous algorithm using a novel flow network-based optimization method. To the best of our knowledge, our scheduling algorithm in a cloud environment, which ensures optimality for real-time tasks and achieves energy savings using dynamic power management simultaneously, is the first in the problem domain. We show that the proposed scheduling algorithm guarantees an optimal schedule for real-time tasks and achieves energy savings simultaneously. Our experimental results show that the proposed algorithm outperforms the latest existing algorithms in terms of both time complexity and energy efficiency.

The energy management for embedded real-time systems is crucial due to their restricted power supplies. With the advancement of technologies, the static energy consumption of the embedded systems that is caused by their leakage power is growing. Thus, a number of research works have started focusing on reducing the static energy consumption by making the systems transit into low-power states, wherein some hardware components are temporarily shut down. Specifically, when a processor is idling, they attempt to set the processor into one of several low-power states. To make a processor remain in the low-power state as long as possible to minimize the energy consumption, the idle time should be maximally clustered. At the same time, in order to satisfy the real-time constraints of the tasks, the length of the clustered idle time should be estimated accurately. To achieve our objective, we propose energy-efficient real-time scheduling algorithms on symmetric homogeneous multiprocessors with a dynamic power management scheme for periodic real-time tasks. The proposed algorithms rely on a flow network model that effectively helps to cluster the idle time while respecting the real-time constraints. In our experimental evaluation, the proposed algorithms consume a comparable static energy to an existing off-line scheme that is the only suitable existing algorithm in the problem domain. Furthermore, we show that the proposed algorithms consume less static energy than the existing one in a case where the total workload of the given task set is low and the task completion is earlier than expected.

Data augmentation is a well-known technique used for improving the generalization performance of modern neural networks. After the success of several traditional random data augmentation for images (including flipping, translation, or rotation), a recent surge of interest in implicit data augmentation techniques occurs to complement random data augmentation techniques. Implicit data augmentation augments training samples in feature space, rather than in pixel space, resulting in the generation of semantically meaningful data. Several techniques on implicit data augmentation have been introduced for classification tasks. However, few approaches have been introduced for regression tasks with continuous/structured labels, such as object pose estimation. Hence, we are motivated to propose a method for implicit semantic data augmentation for hand pose estimation. By considering semantic distances of hand poses, the proposed method implicitly generates extra training samples in feature space. We propose two additional techniques to improve the performance of this augmentation: metric learning and hand-dependent augmentation. Metric learning aims to learn feature representations to reflect the semantic distance of data. For hand pose estimation, the distribution of augmented hand poses can be regulated by managing the distribution of feature representations. Meanwhile, hand-dependent augmentation is specifically designed for hand pose estimation to prevent semantically meaningless hand poses from being generated (e.g., hands generated by simple interpolation between both hands). Further, we demonstrate the effectiveness of the proposed technique using two well-known hand pose datasets: STB and RHD.